close

Вход

Забыли?

вход по аккаунту

?

A Study of RF/Microwave Components Using Fused Deposition Modeling and Micro-Dispensing

код для вставкиСкачать
AStudyofRF/MicrowaveComponentsUsing
FusedDepositionModelingandMicro-Dispensing
by
JoshuaA.Stephenson
Athesissubmittedinpartialfulfillment
oftherequirementsforthedegreeof
MasterofScienceinElectricalEngineering
DepartmentofElectricalEngineering
CollegeofEngineering
UniversityofSouthFlorida
MajorProfessor:ThomasWeller,Ph.D.
LawrenceDunleavy,Ph.D.
JingWang,Ph.D.
DateofApproval:
June13,2017
Keywords:3DPrinting,Ku-band,Coupler,
PowerAmplifier,Connector
Copyright©2017,JoshuaA.Stephenson
ProQuest Number: 10601991
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
ProQuest 10601991
Published by ProQuest LLC (2017 ). Copyright of the Dissertation is held by the Author.
All rights reserved.
This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
ACKNOWLEDGMENTS
I would like to thank Dr. Thomas Weller for his continuous support throughout my
academic career. His ability to make time for all his students and provide quality feedback is
commendable. I would also like to thank Dr. Larry Dunleavy for bestowing his knowledge of
multipleRFandMicrowavetopicsonmeinmyacademicjourney.I’dalsoliketothankhimfor
the opportunity that he provided me while interning at Modelithics. I’d like to thank Dr. Jing
Wangforvolunteeringtobeonmycommittee,itwouldn’thavebeenpossiblewithouthim.I’d
liketothankDr.GokhanMumcuforadoorthatwasalwaysopen.Therewasn’tatimewhereI
walkedoutofhisofficewithoutanansweroradirectiontoproceed.
I’d also like to thank everyone at Sciperio Inc. including but not limited to: Casey
Perkowski, Dr. Paul Deffenbaugh, and Dr. Ken Church. Without their continuous support and
dedicationthisthesiswouldn’thavebeenpossible.ItisalsoimportanttothankScottSkidmore
forallthetimehespenthelpingmewithHFSSandothermeasurementrelatedquestions.I’dlike
tothankMerveKacarforherCSTsoftwareknowledge,theworkinthisthesiswouldn’thavebeen
nearlyasgoodwithouther.
The work in Chapter 3 was supported by the US Air Force Research Laboratory under
contract#FA8650-14-C-2421.TheworkinChapter4and5wassupportedbytheAviationand
Missile Research, Development, and Engineering Center under contract #W31P4Q-16-C-0118.
DistributionStatementA–ApprovedforPublicRelease–DistributionUnlimited.
TABLEOFCONTENTS
LISTOFTABLES
LISTOFFIGURES
ABSTRACT
CHAPTER1:INTRODUCTION
1.1ThesisOverviewandContributions
CHAPTER2:DIRECTDIGITALMANUFACTURINGBACKGROUND
2.1Introduction
2.2FusedDepositionModeling
2.3Micro-Dispensing
2.4Conclusion
CHAPTER3:2.45GHZHYBRIDCOUPLER
3.1Introduction
3.2HybridCouplerBackground
3.3ModifiedDirectionalCoupler
3.4PCBVersionSimulationandMeasuredResults
3.5DDMVersionSimulationvsMeasurementResults
3.6Conclusion
CHAPTER4:KU-BANDCONNECTOR
4.1Introduction
4.2DDMConnectorDesign
4.3DDMTestFixtureDesign
4.4DesignConsiderations
4.5SimulationvsMeasurementResults
4.6TransitionModeling
4.7Conclusion
CHAPTER5:DDMTHERMALMANAGEMENT
5.1Introduction
5.2ThermalBackground
5.3ThermalandRFMeasurements
i
iii
iv
ix
1
1
4
4
4
6
7
8
8
9
11
19
24
26
27
27
27
33
42
45
52
55
56
56
56
58
5.4PowerAmplifierThermalStudy
5.5Conclusion
CHAPTER6:CONCLUSION
6.1Summary
6.2RecommendationsforFutureWork
REFERENCES
APPENDIXA:SIMULATIONANDASSEMBLY
A.1Introduction
A.2SimulationIssues
A.3AssemblyandFabricationTechniques
APPENDIXB:DDMCONNECTORPRINTING
B.13DPrintingProcedure
ii
67
77
78
78
79
80
82
82
82
85
87
87
LISTOFTABLES
Table3.1
Table3.2
Table3.3
Table3.4
Table3.5
Table3.6
Table4.1
Table5.1
Table5.2
Designrequirementsforthe3dBhybridcoupler
12
MaterialpropertiesoftheRogers4725JXRlaminate
13
Dimensionsandcharacteristicimpedanceofthe3dBringhybrid
13
Designrequirementsandachievedperformanceoftheringhybrid
13
MaterialpropertiesoftheRogers4003Claminate
20
Materialpropertiesof100%in-fillABS
24
50-ohmcoaxiallineandDDMconnectordimensionsandparameters
28
Temperaturedataofthetwodesignswhileoperatedovervariousbias
conditions
61
MaterialpropertiesofthevariouscomponentsofthePAtestfixture
68
iii
LISTOFFIGURES
Figure2.1
Figure2.2
Figure2.3
Figure3.1
Figure3.2
Figure3.3
Figure3.4
Figure3.5
Figure3.6
Figure3.7
Figure3.8
Figure3.9
Figure3.10
Figure3.11
Figure3.12
Figure3.13
Figure3.14
Figure3.15
Basicdesktop3Dprintersetup
5
3Dprinterdesignsettings
6
Micro-dispensingsystemusedtoprintconductiveink
7
4-portnetworkforadirectionalcoupler
9
3dBhybridringcoupler
12
Simulatedvsmeasuredresultsofthe3dBhybridringcoupler
13
λ/4transmissionlineusedinthehybridcoupler
14
λ/4equivalentcircuit
15
3λ/4transmissionlineusedinthehybridcoupler
16
3λ/4equivalentcircuit
16
Miniaturizedringhybridwithparallelinductorsandcapacitors
17
Simulatedresultsoftheideal3dBmodifiedhybrid
17
Miniaturizedringhybridwithoutparallelinductorsandcapacitors
18
Simulationresultsoftheidealhybridwithoutparallelcomponents
19
Simulationresultsofthemodifiedidealhybridwithoutparallelelements
19
HybridcouplerwithmodifiedmicrostriplinesandModelithicsmodels
20
Simulationresultsofthemodifiedhybridwithrealisticelements
21
Hybridcouplerwiththeisolatedportterminatedin49.9Ω
21
iv
Figure3.16
Figure3.17
Figure3.18
Figure3.19
Figure3.20
Figure4.1
Figure4.2
Figure4.3
Figure4.4
Figure4.5
Figure4.6
Figure4.7
Figure4.8
Figure4.9
Figure4.10
Figure4.11
Figure4.12
Figure4.13
Figure4.14
Figure4.15
Comparisonofthemodifiedhybridwithandwithouta49.9Ω
terminatedisolationport
22
LayoutandpictureoftheRogers4003Chybridcoupler
22
Simulatedvsmeasurementdataforthefabricatedcoupler
23
LayoutandpictureoftheDDMhybridcoupler
24
SimulatedvsmeasurementdataforthefabricatedDDMcoupler
25
KonnectRFKAD178244adapterusedasaninterfacebetweenanSMA
assemblyandtheDDMconnector
28
Simulationmodelandresultsvaryingtheviaoffset
30
Simulationmodelandresultsvaryingthetoptaper
31
S-parametersofthetaperedconnectordesign
32
SimulationmodeloftheDDMtestfixture
33
S-parametersoftheinitialteststructuredesign
34
TDRplotofchangesintaperwidth
35
Simulationresultsandmodelvaryingtheinsetradius
36
S-parametersoftheDDMteststructurewithaninsetradiusof1mm
37
ElectricfieldconfigurationinsidetheDDMconnectorattheDDM-
microstrip
37
DDMmodelwiththesemi-circularconetransitionaddedwithinthe
structure
38
DDMtestfixturesimulationresultsvaryingthetransitionradius
38
DDMtestfixturesimulationresultsvaryingthetransitionheight
39
ElectricfieldconfigurationoftheDDMconnectorwiththeoptimized
transition
41
S-parametersofthestructurewithdifferentdielectricconstants
41
v
Figure4.16
Figure4.17
Figure4.18
Figure4.19
Figure4.20
Figure4.21
Figure4.22
Figure4.23
Figure4.24
Figure4.25
Figure4.26
Figure4.27
Figure4.28
Figure4.29
Figure4.30
Figure4.31
Figure4.32
Figure4.33
Differentprintingorientationstodemonstratetheissueswith
printingtheDDMconnectortestfixture
42
Modificationrequiredtosuccessfullyprinttheteststructure
42
S-parametersoftheteststructurewiththe0.1mmprintlayeradded
43
S-parametersoftheteststructurewiththemodifiedviaoffset
44
Modelandsimulationdataoftheteststructurewithandwithoutthe
mechanicalstructure
44
3DmodelandfabricatedDDMtestfixture
45
MeasuredvssimulatedS-parameterdataoftheteststructure
46
Fabricatedand3Dmodelofdamagedtransition
47
S-parametersweepoftheABStransitionheightwithoneconnector
damaged
48
S-parametersweepofthegapradiuswithoneconnectordamaged
49
S-parametersweepoftheABStransitionheightwithbothconnectors
damaged
50
S-parametersweepofthegapradiuswithbothconnectorsdamaged
50
S-parameterdataoftheDDMtestfixturewithadamagedconnector
51
Lumpedcomponentmodelforatransmissionline
52
Lumpedcomponentmodeloftheteststructurewithoutthemodified
transition
53
Performanceofthelumpedcomponentmodelwithoutamodified
transition
53
Lumpedcomponentmodeloftheteststructurewiththemodified
transition
54
Performanceofthelumpedcomponentmodelwithamodified
transition
54
vi
Figure5.1
Figure5.2
Figure5.3
Figure5.4
Figure5.5
Figure5.6
Figure5.7
Figure5.8
Figure5.9
Figure5.10
Figure5.11
Figure5.12
Figure5.13
Figure5.14
Figure5.15
Figure5.16
Figure5.17
Figure5.18
Figure5.19
ConvectioncoefficientvstemperatureusedinAnsysWorkbench
57
ModifiedtestboardusedforanABSsubstrate
58
Designswithvaryingvianumbers
59
FabricateddesignwithacarrierandSouthwestMicrowaveconnectors
attached
59
MeasurementsetupforthePA
60
Imagesofthe16viadesignatbiascondition5
61
Temperaturevsoutputpowerforthetwodesigns
63
Outputpowerandgainvsinputpowerforthetwodesigns
64
ThermalimageandtemperatureprofilewiththePAon
66
ThermalimageandtemperatureprofilewiththePAoff
66
BasictopologyoftheQFNpackagemountedonasubstrateandcarrier
67
AnsysWorkbenchviewofthesimulationmodel
68
QFNstackupandtheassociateddimensions
69
25viaholedesignwitha1.5Wheatsourceappliedtothetopfaceof
thedie
70
16viaholedesignwitha1.5Wheatsourceappliedtothetopfaceof
thedie
70
PApadlayoutwithvaryingepoxyareas
71
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
4mm2epoxyarea
72
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
2.25mm2epoxyarea
72
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
1mm2epoxyarea
72
vii
Figure5.20
Figure5.21
Figure5.22
Figure5.23
Figure5.24
Figure5.25
Figure5.26
FigureA.1
FigureA.2
FigureA.3
FigureA.4
FigureA.5
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
thermalconductivityof29W/mK
73
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
thermalconductivityof15W/mK
74
16viaholedesignwith1.5Wappliedtothetopfaceofthedieanda
thermalconductivityof5W/mK
74
Maximumtemperaturevspowerdissipatedforthe25viadesign
75
25viadesignwith3.5Wappliedtothetopfaceofthedie
76
16viadesignwith3.5Wappliedtothetopfaceofthedie
76
25viadesignwith1.5Wappliedtothetopfaceofthediewithout
acarrier
76
EMsetupoptionstoavoidsimplifyingthelayoutmesh
83
Meshresultingfromthe“Simplifythelayout”option
83
Meshresultingfromnotselectingthe“Simplifythelayout”option
84
Meshresultingfromselectingthe“Meshreduction”option
84
TopABSlayerseparatingfromthegroundplane
86
viii
ABSTRACT
The design and study of multiple RF direct digital manufactured (DDM) devices are
presentedinthiswork.A2.45GHz,180°hybridcouplerisdesignedtoprovidethespacerequired
for other system components. The coupler is designed and manufactured on a 32 mil Rogers
4003Csubstrateandadaptedtoa100%in-fillacrylonitrilebutadienestyrene(ABS)substrate.A
sizereductionof66%isaccomplishedwithabandwidthof16%.ADDMKubandconnectoris
modeledandfabricatedusingvaryingrelativedielectricconstantsof50%and100%in-fillABS.
Theconnectormaintainslessthan0.45dBofinsertionlossupto14GHzandgreaterthan10dB
ofreturnlossupto15GHz.Alumpedcomponentmodelisalsodevelopedtomodelthedamaged
transitionoftheconnectorwithagreementtonumericalelectromagneticsimulationsoftware.
Lastly,athermalandRFstudyofaKubandpoweramplifier(PA)isperformed.Two5mil100%
in-fillABSPAtestfixturesarefabricatedwithavaryingnumberofvias.Thedesignsarebiasedat
variousoperatingpointstocollectthermalandRFdata.ThePAoperatesat151°Cbeforemelting
the ABS substrate. A thermal model is developed from the measurement data to predict the
temperaturesatgivenpowerlevelswithgoodagreementbetweensimulationandmodeldata.
ix
CHAPTER1:INTRODUCTION
3Dprintingisawell-knowntermfromhobbyiststoscientistsandengineers.However,
notsocommonknowledgeisthat3Dprintingandrapidprototyping(RP)havebeenaroundsince
thelate1980’s.ThefirstRPtechnologywasthestereolithography(SL)machinefirstpatentedby
CharlesHullin1984.TheSLmachineusesaUVcurableliquidandafocusedUVlighttocreate3D
objectslayerbylayer[1].In1989,thefamousfuseddepositionmodeling(FDM)technologywas
inventedbyScottCrump[2].Sincethen,thepatenthasexpiredandmanyindustrieshavemoved
in to take advantage of the technology. From aerospace to biological devices, 3D printing is
applicableinmanyfields.3Dprintingisn’tareplacementfortraditionalhighyieldmanufacturing
processes. However, there is a clear benefit of 3D printing when manufacturing low volume
highlycustomizedpartsorforrapidprototypingofaconceptdesigntolaterbuildwithtraditional
manufacturingmethods.
1.1ThesisOverviewandContributions
Thepurposeofthisthesisistheuseofdirectdigitalmanufacturing(DDM)todesignRF
andmicrowavedevices.3DprintingRFdevicespresentsmanychallengesduetothestringent
requirements on line dimensions, relative dielectric constants, conductivity and thermal
properties. When operating at higher frequencies parts reduce in size and this increases
manufacturingdifficulty.Tolimitthesechallenges,ahighprecisionprinterisneeded.ThenScrypt
3DnseriesprinterwithnFDandSmartPumpTechnologiesenablesthehighprecisionextrusion
1
anddispensingofthermoplasticsandconductivepastes,respectively[3,4].Thethermoplastic
used in this work is acrylonitrile butadiene styrene (ABS). Also used are DuPont CB028 silver
conductor[5]whichismicro-dispensedusingtheSmartPumpandEpoxyTechnologyH20Eepoxy
[6]isusedforassemblypurposes.Chapter2willdiscussthebasicbackgroundof3Dprintingand
thetechnologiesused.
Whenconsideringadesign,thecostperareaisamaincontributortodesigndecisions.
Additionally,whenspaceisn’tavailableit’simportantthatdesignersfindwaysofminiaturizing
componentstomeetdesignspecificationsandpackagingrequirements.Thesizereductionofa
2.45GHz,180°hybridcoupleristhefirstmajorcontributionofthisthesis,aspresentedinChapter
3. The miniaturization of the hybrid coupler is accomplished by using capacitively loaded
transmissionlines.Thisreducesthesizeofthetransmissionlineswhilemaintainingthephase
characteristics.
The second major contribution of this thesis is the demonstration of a Ku band DDM
connector. Chapter 4 examines the utility of the time domain reflectometry (TDR) tool in
designingtheconnector.Multipleparametersweepsareperformedtoprovideinsightintothe
contributingfactorsofimpedancemismatchesatdiscontinuities.Also,anembeddedsemicircular
transitionisdevelopedtoconverttheimpedancesandelectromagneticfieldconfigurationsofa
coaxialconnectortoamicrostripline.
Chapter 5 details the last contribution of this thesis, with the study of the thermal
performanceofaKu-bandpoweramplifier(PA)anditseffectonRFperformance.Asimulation
modelforthetestfixtureisalsodevelopedthatcanbeusedtodetermineifagivenmaterialwill
sufficientlydiffusetheheatawayfromthePAchip.
2
Chapter6willconcludethethesiswithasummaryoftheresearchandrecommendations
for future work. Appendix A provides information about assembly techniques used as well as
specialconsiderationswhenusingsimulationsoftware.AppendixBprovidesinformationabout
theprintingprocedurefortheDDMconnector.
3
CHAPTER2:DIRECTDIGITALMANUFACTURINGBACKGROUND
2.1Introduction
Therearemultipleformsofdirectdigitalmanufacturing(DDM)withstereolithography
(SL), selective laser sintering (SLM), and fused deposition modeling (FDM) being a few of the
popular technologies [7]. All three technologies fabricate parts from CAD files layer-by-layer,
howevertheprocesswithwhichtheybuildthe3Dpartsvaries.SLusesalaserandphotosensitive
resintobuilditsparts.SLandSLMarealike,butinsteadofaphotosensitiveresinSLMusesa
powderedmaterial.FDM,thetechnologyusedinthiswork,usesathermoplasticfilamentwhich
is melted with an extrusion head. FDM typically builds the model from the bottom, however
dependingonthestructurethiscanvary.Section2.2willintroducetheFDMprocessandthe
requiredcomponents.Section2.3willcovermicro-dispensingofconductivepastesthatareused
mainlyfortransmissionlinesandtofillviaholes.
2.2FusedDepositionModeling
FDMisanextrusionbased3Dprintingtechnology.Figure2.1showsabasicdesktop3D
printersetup.Therearemanydifferentthermoplasticfilamentsavailabletoinclude:acrylonitrile
butadienestyrene(ABS),polylacticacid(PLA)andpolyetherimide(ULTEM).Filamentsaresoldby
thespoolandarefedintothefilamentfeedingsystem.Thefilamentfeedingsystemconsistsof
variouscomponentstopullthefilamentfromthespooltotheheatingelement.Dependingon
4
thetypeofmaterialused,theheatingelementtemperaturewillbeadjustedtomaintainflowof
thematerialontotheheatedprintbed.Fromtherethemodelwillbebuiltfromthebottomup.
Figure2.1Basicdesktop3Dprintersetup
3Dprintingtechnologyprovidesadesignerwithmanyoptionstocustomizedesigns.One
suchcustomizationisin-fillpercentage.In-fillpercentageistheamountofmaterialwhichwillbe
used to fill the inside of the model. Figure 2.2a shows three different in-fill percentages of a
rectilinearfillpattern.Thefillpatternisexactlywhatitsoundslike,thepatternthatwillbeused
tofilltheinsideofthemodel.Figure2.2bshowsthethreefillpatternswhicharejustafewof
manytypes.Thefillpatternwillmaintainthein-fillpercentageregardlessofthepatternchosen.
Byvaryingthein-fillpercentage,thedielectricconstantcanbechangedtoaccommodatethe
design.Both50%and100%in-fillpercentagesareusedinthiswork,witharectilinearpattern
(Figure2.2a).
5
(a)
(b)
Figure2.23Dprinterdesignsettings.(a)varyingin-fillpercentage(b)varyingin-fillpattern
The 3D printing system used in the presented research is the nScrypt 3Dn series. The
nScryptsystemutilizesafilamentfeedingsystemcalledthenFD.ThenFDmovementisrestricted
to the z-axis only. Where the heated print bed restricts movement to the x-y axis. The x-y
resolutionofthenScryptprinteris10nm–1μmandinthezdirectionis0.5μm[8].Thishigh
resolutionprovidestheprecisionneededtoprintqualityRFandmicrowavedevices.
2.3Micro-Dispensing
Micro-dispensingisatechnologythatenablestheusertoprintavarietyofmaterialswith
varyingviscositiesincludingepoxiesandconductiveinkswithhighprecision.Theabilitytoprint
6
linesassmallas25μmispossiblewithnScryptSmartPumptechnology[4].Theconfigurationof
themicro-dispensingsystemisshowninFigure2.3.Conductiveink,e.g.DuPontCB028,isloaded
intoasyringeandconnectedtotheSmartPump.Acomputercontrolsthedispensingaccording
toaninputfileoftheelectroniccircuittheuserwouldliketoprint.
Figure2.3Micro-dispensingsystemusedtoprintconductiveink
2.4Conclusion
FDMandmicro-dispensingalonearepowerfultechnologies,buttogetherthepossibilities
areendless.3Dprintingprovidesthedesignerwithaddedflexibilityinthetypesofstructures
thatarepossible.Varyingthein-fillpercentageallowsthedielectricconstanttobevariedlayerby-layer.Multiplelayersofthermoplasticsandconductiveinksprovideflexibilityandalow-cost
solutiontotraditionalmanufacturingtechnologies.Lastly,itprovidestheabilitytoprintmultimaterialcircuitsanddecreasesthetimetoafirstpassdesign.
7
CHAPTER3:2.45GHZHYBRIDCOUPLER
3.1Introduction
Couplersarenotnewtomicrowaveengineering.Theyhavebeenusedinvariousdesigns
throughouthistorytoinclude:baluns,mixers,andamplifiers[9,10].Thetopologiesofhybrids
provideeither0°,90°or180°phasedifferenceattheoutputports.Thefocusofthisresearchis
inaDDMfabricated180°hybridcouplerthatisusedtofeedthetwoportsofacircularlypolarized
antenna.Thecoupleroperatingfrequencyis2.45GHzandrequiresa180±5°phasedifference
betweenthecoupledandthroughports.Themaindrivingrequirementofthedesignisthesize
ofthecoupler,whichneedstobelessthan736mm2.Thesmallsizewillprovidethespaceneeded
forothersystemcomponentsonaphasedarrayunitcell.Insection3.2thegeneraltheorybehind
thehybridcouplerwillbepresented.Abriefderivationwillbediscussedtoprovideabasisforits
operation.AvarietyoftechniquestoreducethesizeofthecouplerwillbediscussedinSection
3.3.AlthoughthecouplerwillbefabricatedusingDDM,itisimportanttofabricatethecoupler
using traditional (subtractive) printed circuit board (PCB) manufacturing techniques for
comparison.Section3.4isadiscussiononthesimulationandmeasurementresultsofthePCB
versionofthecoupler.Section3.5providesthesimulationversusmeasuredresultsoftheDDM
coupler.
8
3.2HybridCouplerBackground
Figure3.14-portnetworkforadirectionalcoupler
Directionalcouplersare4-portdevices(Figure3.1)thatcanbereciprocal,matchedatall
ports and lossless under specific conditions. The derivation of the directional coupler, as
discussedin[11],willbepresentedbelow.TheS-parametermatrixofa4-portnetworkcanbe
simplifiedifit’sreciprocal(Sij=Sji):
S## S S = #%
S#& S#' S#% S%% S%& S%' S#& S%& S&& S&' S#' S%' S&' S''
(3.1)
andmatchedatallports,resultinginadiagonalmatrix(Sij=0wherei=j):
0
S S = #%
S#& S#' S#% 0
S%& S%' S#& S%& 0
S&' S#' S%' S&' 0
(3.2)
Ifthenetworkislosslessitneedstosatisfytheunitarypropertiesinequations3.3and3.4.
*
).# S)* S)+
= 0, fori ≠ j
*
).# S)* S)*
= 1
(3.3)
(3.4)
Applyingequation3.3tothematrixin3.2resultsinequations3.5-3.6below:
*
S#'
S#& % - S%'
%
= 0
(3.5)
S%& S#% % - S&'
%
= 0
(3.6)
9
Onesolutionto3.5and3.6isthatS14andS23=0.ThissolutionfurthersimplifiestheS-parameter
matrixof3.2,shownbelow:
0
S#%
S =
S#&
0
S#%
0
0
S%'
S#&
0
0
S&'
0
S%'
S&'
0
(3.7)
Applyingequation1.4tothematrixin1.7,resultsinequations1.8-1.11:
S#%
%
+ S#&
%
= 1
(3.8)
S#%
%
+ S%'
%
= 1
(3.9)
S#&
%
+ S&'
%
= 1
(3.10)
S%'
%
+ S&'
%
= 1
(3.11)
The following relationships emerge when solving the system of equations: |S13| =
|S24|and |S12| = |S34|. After further simplification and selection of the phase constants, the
followingmatrixresults:
0
α
S =
β
0
α
0
0
-β
β
0
0
α
0
-β
α
0
(3.12)
whereαandβarerealconstants.Theabovematrixisalsoknownasanantisymmetriccoupler
duetothe180°phasedifferencebetweenS13andS24.Theabovematrixisthebasisforthehybrid
couplerandwillbeusedinthedesignprocess.
What the matrix in 3.12 describes is that depending on the input port(s) chosen, the
designer can combine or divide the input power between a port(s) with a phase shift that
dependsonthephasereferenceoftheoutputport(s).Forexample,ifport1ischosentobethe
inputport,thepowerwillbesplitbetweenports2and3,andbothoutputportswillbeinphase.
10
However,iftheinputportischosentobeport4,thepowerwillbesplitbetweenports2and3,
andthetwosignalswillbe180°outofphase.
Todeterminehowthepowerisdividedthedesignercanchangethecouplingfactorand
utilizetheconservationofpowertodeterminethepowersplitbetweenports,equations3.13
and3.14,respectively.
C = −20log β dB
α%+ β
%
= 1
(3.13)
(3.14)
Thecouplingfactorusedforthisapplicationwillbe3dB.Utilizingthetwoequationsabove,α=
β=0.5.Matrix3.12canthenbesimplifiedto:
S =
+
%
0 1
1 0
1 0
0 −1
1 0
0 −1
0 1
1 0
(3.15)
Thematrixin3.15showsthatwhenanexcitationisappliedtoanyoneport,theresultingoutputs
willbehalfthepowerandeitherinphaseoroutofphase,dependingontheinputport.
3.3ModifiedDirectionalCoupler
Apopulartopologyfora180°hybridcoupleristheringhybridorratrace(Figure3.2).The
ring hybrid has the behavior of the S-parameter matrix in 3.15. This section will cover the
topologyandsimulationinformationforthe3dBringhybridandthemodifiedringhybrid.
11
(a)
(b)
Figure3.23dBhybridringcoupler.(a)layout(b)fabricated
Table3.1Designrequirementsforthe3dBhybridcoupler
Parameter
LayoutSize
Coupling
ReturnLoss
PhaseDifference(°)
Bandwidth
Requirements
<736mm2
<4dB
>10dB
180°±5°
15%
The design requirements for the coupler are in Table 3.1. Keysight Advanced Design
System(ADS)isusedtorealizethecircuitinFigure3.2.Thesubstrateusedforthissimulationis
theRogers4725JXR,whosematerialpropertiescanbefoundinTable3.2[12].ADSLinecalcis
usedtodeterminethewidthsandlengthsofthevariousmicrostriplines.Thesamedimensions
and lengths can be calculated/verified using the various equations in [11]. The resulting
microstripdimensionsareshowninTable3.3.
12
Table3.2MaterialpropertiesoftheRogers4725JXRlaminate
Parameter
DielectricConstant(Er)
SubstrateHeight(H)
ConductorThickness(T)
Conductivity(κ)
LossTangent(TanD)
Value
2.64
30.7mil
25μm
5.8x107Sm-1
0.002
Table3.3Dimensionsandcharacteristicimpedanceofthe3dBringhybrid
MicrostripLine
λ/4
3λ/2
Ports
Z0 70.71Ω
70.71Ω
50Ω
Width
1.11mm
1.11mm
2.09mm
ArcLength
21.21mm
63.62mm
9.47mm
(a)
(b)
Figure3.3Simulatedvsmeasuredresultsofthe3dBhybridringcoupler.
(a)S44,S24,S34(b)Unwrappedphasedifferencebetweenoutputports
Solid–SimulatedDashed–Measured
Table3.4Designrequirementsandachievedperformanceoftheringhybrid
Parameter
LayoutSize
Coupling
ReturnLoss
PhaseDifference(°)
Bandwidth
Requirements
<736mm2
<4dB
>10dB
180°±5°
15%
13
Achieved
3721mm2
<4dB
>15dB
180°±5°
16%
The 3dB hybrid ring coupler is simulated using an ADS momentum simulation. The
simulated vs measurement results for the circuit in Figure 3.2 are shown in Figure 3.3. The
simulation shows that the power division between the two outputs is 3 dB. Whereas, the
measurementshaveanoutputpowerof3.5dBatthecenterfrequency(Figure3.3a).Although,
theoveralltrendofthepowersplitshowsagreement.Figure3.3bshowstheunwrappedphase
differencebetweenthetwooutputports.Thephasedifferenceofthesimulatedandmeasured
circuitis179.8°and182°atthecenterfrequency,respectively.Table3.4summarizestheresults
anddesigngoals.
Althoughtheratracecouplerprovidessufficientperformancethesizeofthecoupleris
toolarge.Tomeetthisrequirement,thecircuitwillneedtobemodifiedtoreducethesizewhile
maintainingtheRFperformance.Thiswillbeaccomplishedthroughminiaturizationtechniques
suchascapacitiveloading[10].Bycapacitiveloadingthetransmissionline(TL),thedesignerwill,
in effect, reduce the length while maintaining the same phase delay. Figure 3.4 shows the
topologyofaλ/4transmissionlineandtheassociatedABCDmatrix.Theequivalentcircuitand
associatedABCDmatrixusedtominiaturizetheλ/4transmissionlineisshowninFigure3.5.
AC
CC
'
'
BC
DC
(a)
'
=
'
0
jZC
jYC
0
'
'
(b)
Figure3.4λ/4transmissionlineusedinthehybridcoupler.
(a)equivalentcircuit(b)ABCDmatrix
14
AG
CG
(a)
+ jYG ZJK sin βlJK
cos βlJK
BG
=
DG
2YG cos βlJK + j sin βlJK
jZJK sin βlJK
YJK + YG% ZJK jYG ZJK sin βlJK + cos βlJK
(b)
Figure3.5λ/4equivalentcircuit.(a)networktopology(b)ABCDmatrix
For the two transmission lines to exhibit the same RF behavior the ABCD matrix of the two
networksmustbeequal:
cos βlJK + jZJK YG sin βlJK = 0
(3.16)
jZJK sin βlJK = jZC (3.17)
2YG cos βlJK + jYJK sin βlJK + jYG% ZJK sin βlJK = jYC (3.18)
jYG ZJK sin βlJK + cos βlJK = 0
(3.19)
'
'
Solvingequations3.16-3.19forZEQ,Ya,andlEQ,resultsinthefollowingrelationships:
C
lJK = (3.20)
ZJK = 2ZN (3.21)
CG =
M
#
%OP QP
(3.22)
ThesolutionstotheequivalentnetworkinFigure3.5resultinatransmissionlinewhich
ishalfofthesizeofthepreviousratracecouplerTL.Thisispossibleduetotheshuntcapacitances
definedbyEquation3.22.Asimilarprocedureisusedforthe3λ/4TLoftheringhybridinFigure
3.2.Figure3.6showsthe3λ/4TLandtherespectiveABCDmatrix.Theequivalentpinetworkthat
15
canbeimplementedtominimizethephysicallengthofthetransmissionlineisshowninFigure
3.7,alongwithitsABCDmatrix.
A&C
C&C
'
'
B&C
D&C
(a)
'
=
'
0
-jZ&C
-jY&C
'
0
'
(b)
Figure3.63λ/4transmissionlineusedinthehybridcoupler.
(a)equivalentcircuit(b)ABCDmatrix
(a)
1 + YZR
ZR
A R BR
=
%
CR DR
2Y + Y ZR YZR + 1
(b)
Figure3.73λ/4equivalentcircuit.(a)networktopology(b)ABCDmatrix
Afterequatingthetwomatrices(3.6band3.7b)andsolvingtheresultingequationsthefollowing
relationsareobtained:
L =
CR =
%QP
OP
#
%OP QP
(3.23)
(3.24)
Usingequations3.20-3.24,andtheequivalentnetworksfortheλ/4and3λ/4theminiaturized
ringhybridisdetermined,asshowninFigure3.8.
16
Figure3.8Miniaturizedringhybridwithparallelinductorsandcapacitors
ThecircuittopologyofFigure3.8issimulatedinanADSschematicusingidealcomponents
(capacitors,inductors,TL).ThecouplerissimulatedusingADSandtheresultsareshowninFigure
3.9.Thecapacitances,inductancesandtransmissionlineproperties/valuesarecalculatedusing
theequationsderivedabove.Figure3.9ashowsthatthepowerisequallysplitwiththeinsertion
losses being approximately 3 dB at the design frequency of 2.45 GHz. The phase difference
betweentheoutputportsiswithin±5°acrosstheentirebandwidth(Figure3.9b).
(a)
(b)
Figure3.9Simulatedresultsoftheideal3dBmodifiedhybrid.(a)InsertionandReturnLoss(b)
Unwrappedphasedifferencebetweenoutputports
17
TheimpedanceoftheoftheparallelLCishighatthedesignfrequency,allowingforthe
circuitinFigure3.8tobesimplified,byremovingthetwocomponents.Thelayoutoftheresulting
circuitisshowninFigure3.10.
Figure3.10Miniaturizedringhybridwithoutparallelinductorsandcapacitors
The simulation results of the ADS schematic in Figure 3.10 are shown in Figure 3.11.
Removing the parallel LC, changes the input return loss of the hybrid as well as the phase
difference between the direct and coupled ports. The change in the phase difference can be
compensatedbychangingthevalueofCato1.5pFandthelengthsofthetransmissionlinesto
47°(Figure3.10).ThesimulationresultsofthemodifiedhybridareshownFigure3.12.Removing
theextracomponentswillreducetheoverallfootprintoftheoverallcoupler,howeverthephase
performance degrades. This performance degradation can be addressed when parasitic
components,interconnectsandtransmissionlineelementsareintroduced.
18
(a)
(b)
Figure3.11Simulationresultsoftheidealhybridwithoutparallelcomponents.
(a)InsertionandReturnLoss(b)Unwrappedphasedifferencebetweenoutputports
(a)
(b)
Figure3.12Simulationresultsofthemodifiedidealhybridwithoutparallelelements.
(a)InsertionandReturnLoss(b)Unwrappedphasedifferencebetweenoutputports
3.4PCBVersionSimulationandMeasuredResults
Theprevioussectionsprovidedthefoundationforthedesignofareducedsizeringhybrid.
In this section, the parasitic design, simulation, and measurement results are presented. The
substrateusedinthedevelopmentofthedesignisa32milRogers4003Claminate[13],whose
substratepropertiesaresummarizedinTable3.4.Aλ/8transmissionline,at2.45GHz,hasthe
following properties: length of 9.88 mm, width of 0.44 mm and an impedance of 100 Ω. The
coupler in Figure 3.10 is physically unrealizable due to the transmission line elements. The
straighttransmissionlinesbetweentheinputandcoupledports,anddirectandisolationports
19
willnotbeconnectedintheformofFigure3.10.Thisissueissolvedusingcurvedtransmission
lineelementsasshowninFigure3.13.ThetunedcircuitschematicandlayoutareshowninFigure
3.13aandFigure3.13b,respectively.Whentheexactdimensionsandcapacitorvaluesareused
from Figure 3.10 the performance of the circuits is degraded. The curved transmission line
lengthsandcapacitorvaluesneedtobereducedwithCa,CbandLλ/4changedto:1pF,0.7pF,and
9.48mm,respectively.ThesimulatedresultsofthetunedcircuitareshowninFigure3.14.
Table3.5MaterialpropertiesoftheRogers4003Claminate
Parameter
DielectricConstant(Er)
SubstrateHeight(H)
ConductorThickness(T)
Conductivity(κ)
LossTangent(TanD)
Value
3.38
32mil
35μm
5.8x107Sm-1
0.002
(a)
(b)
Figure3.13HybridcouplerwithmodifiedmicrostriplinesandModelithicsmodels.
(a)Schematic(b)Layoutwithcouplerdimensions
20
(a)
(b)
Figure3.14Simulationresultsofthemodifiedhybridwithrealisticelements.
(a)InsertionandReturnLoss(b)Unwrappedphasedifferencebetweenoutputports
The hybrid in Figure 3.13 has a coupling value, at 2.45 GHz, of 3.1 dB and a phase
difference of 181°. The coupling value remains within 3 ±1 dB across the bandwidth of the
coupler.
(a)
(b)
Figure3.15Hybridcouplerwiththeisolatedportterminatedin49.9Ω.
(a)Schematic(b)Layoutwithcouplerdimensions
21
(a)
(b)
Figure3.16Comparisonofthemodifiedhybridwithandwithouta49.9Ωterminatedisolation
port.(a)InsertionandReturnLoss(b)Unwrappedphasedifferencebetweenoutputports
Solid–WithoutTerminationDashed–WithTermination
Thesimulationdataofthe4-porthybridmeetsallthedesignrequirementsofTable3.1.
However, the coupler’s application requires that the isolation port be terminated in 50 Ω, as
showninFigure3.15.ThesimulationresultsinFigure3.16showthatthechangesinthecoupling,
returnlossandphasedifferencebetweenthe4-porthybridandtheterminatedisolationport
hybridareminimal.
(a)
(b)
Figure3.17LayoutandpictureoftheRogers4003Chybridcoupler.(a)Layoutwith
dimensionsofthePCB.(b)Pictureofthefabricatedhybridcoupler.
Tofabricatethecircuit,feedlinesareaddedtothecoupler.Thelayoutandthefabricated
couplerdesignareshowninFigures3.17aand3.17b,respectively.TheADSco-simulationand
measureddataareshowninFigure3.18.
22
(a)
(b)
(c)
Figure3.18Simulatedvsmeasurementdataforthefabricatedcoupler.
(a)InsertionLoss(b)ReturnLoss(c)Unwrappedphasedifferencebetweenoutputports
Solid–SimulatedDashed–Measured
There is good agreement between the simulated and measured data. The crossover
frequencyfortheinsertionlossesisapproximately30MHzdifferent(Figure3.17a).However,the
3dBsplitismaintainedatthefrequencyofinterest.Thereturnlossofallportsisgreaterthan15
dB. The area of the Roger 4003C coupler is 231.35 mm2 providing a substantial decrease
comparedtotheratracecouplerofSection3.3.
23
3.5DDMVersionSimulationvsMeasurementResults
The DDM version of the coupler is much like the Rogers 4003C version. The main
differenceisinthewidthofthelinestomaintainthe100ohmimpedancecausedbytherelative
dielectricchanges(Table3.6).
Table3.6Materialpropertiesof100%in-fillABS
Parameter
DielectricConstant(Er)
SubstrateHeight(H)
ConductorThickness(T)
Conductivity(κ)
LossTangent(TanD)
Value
2.423
32mil
25μm
1.65x106Sm-1
0.006
ThelayoutandthefabricatedcouplerdesignareshowninFigures3.19aand3.19b,respectively.
Theco-simulatedandmeasureddataareshowninFigure3.20.Thecouplerareais248.82mm2
which is 66% smaller than the requirement of 736 mm2. The capacitor values did not change
betweenthePCBversionandDDMversion(Ca=1pFandCb=0.7pF).
(a)
(b)
Figure3.19LayoutandpictureoftheDDMhybridcoupler.
(a)LayoutwithdimensionsoftheDDMcoupler.(b)Pictureofthefabricatedhybridcoupler.
24
(a)
(b)
(c)
Figure3.20SimulatedvsmeasurementdataforthefabricatedDDMcoupler.
(a)Insertionloss(b)Returnloss(c)Unwrappedphasedifferencebetweenoutputports
Solid–SimulatedDashed–Measured
TheDDMcouplerhasgoodinsertionandreturnlossesatthedesignfrequencyandthe
measurement to simulated results share the same trend. The insertion loss bandwidth is
approximately16%.Thereturnlossesmaintain10dBandgreaterupto2.9GHz.Themeasured
phasedifferenceofthecouplerisbelowthe180°.Infuturedesigns,thetransmissionlinesshould
betunedtocompensateforthislowphasedifference.
25
3.6Conclusion
Inthischapter,thedesignandfabricationofa180°hybridcouplerwasdiscussed.The
generaltheoryprovidedanunderstandingoftheoperationofthecoupler.Theratracecoupler
providedagoodstartingpointforfurtherminimizationtechniques.Usinglumpedcomponents
and equivalent networks provided a means to reduce the size of the hybrid coupler while
maintaining the performance of the coupler. Although, some of the components could be
neglectedtofurtherreducethesize.Thisreductionincomponentsandsizecomesatacostof
symmetryandperformance.But,sizeisthedrivingrequirementtothedesignsotheperformance
degradationisacceptable.BothPCBandDDMversionsofthereducedsizecouplershowedgood
simulationtomeasurementperformance.However,inthefuturetheoutputphasedifference
willneedimprovements.
26
CHAPTER4:KU-BANDCONNECTOR
4.1Introduction
Interestin3DprintingRFdeviceshasgrownoverthepastfewyears.Theabilityto3D
printfilters,powerdividers,antennas,andphaseshiftershasbecomecommonplace[14-17].
However,achallengethatpresentsitselfwhentestingorconnectingthesecomponentstoother
modules, is a robust connector that can be embedded and printed with the design itself. For
example,ifeachcomponentonaunitcellneededtobetested,thedesignerwouldhavetoprint
theunitcellandbuyofftheshelfpartsandassembletheconnectorsasapostprocessingstep.
However,ifa3Dprintedconnectorcouldbeprintedalongwiththestructurealltheuserwould
havetodoisconnectthecoaxialcable.Thiswouldreducethetimetotestingandtheunnecessary
temperatureexposuretocurethesilverepoxies.Inthischapter,thegeneraldesignprocedureof
the DDM connector (Section 4.2), the design of the test structure (Section 4.3), the design
considerations(Section4.4),thesimulatedvsmeasuredresults(Section4.5),andthetransition
modeling(Section4.6)willbecovered.
4.2DDMConnectorDesign
Thesubminiatureversiona(SMA)tosliponadapterusedforthedesignistheKonnectRF
modelnumberKAD178244,whichhasanupperfrequencyof18GHz[18].TheadapterisaSMA
to slip-on adapter. The adapter is cross sectioned to determine the different dimensions that
wouldbeinaccessibleotherwise.Theadapterandacrosssectionoftheadapterareshownin
27
Figure4.1.ThedesignoftheDirectDigitalManufactured(DDM)connectorisconstrainedbythe
dimensionsofthecommercialoff-the-shelf(COTS)adapter(Figure4.1a).TheDDMconnectorwill
alsobereferredtoasareceptacleinthiswork,howevertheyrefertothesamething(Figure
4.2a).Table4.1liststhedimensionsofa50-ohmcoaxiallineaswellastheDDMconnector.The
COTSadapterdimensionsonthemaleendareusedasthestartingpointfortheDDMconnector,
which is designed using Computer Simulation Technology (CST) numerical electromagnetic
simulationsoftware.AllTDRsimulationsareperformedusingCSTandtheS-parameterdatais
verifiedusingAnsysHighFrequencyStructuralSimulator(HFSS).
(a)
(b)
Figure4.1KonnectRFKAD178244adapterusedasaninterfacebetweenanSMAassembly
andtheDDMconnector.(a)cross-section(b)connector
Table4.150-ohmcoaxiallineandDDMconnectordimensionsandparameters
Parameter
OuterDiameter(mm)
InnerDiameter(mm)
DielectricConstant
LossTangent
Length(mm)
Z0(ohm)
50-ohmcoaxialline
4
1.25
2.1
0.001
12.75
48.125
28
DDM
6.2
1.7
2.4
0.006
4.5
50.0
Understanding the limiting factors of the connector is accomplished by modeling the
connectorinCSTandperformingvarioussimulations.Toavoidanoverlycomplicatedmodelof
the COTS adapter, a 50-ohm coaxial line is used in the simulations as outlined in Table 4.1.
Similarly,theDDMreceptacleismodeledusingthetabulateddataofTable4.1.
Thetimedomainreflectometry(TDR)plotoftheconnectorprovidesinsightintochanges
inimpedancealongthestructureandthedistanceatwhichdiscontinuitiesoccur.Equation4.1is
usedtounderstandthediscontinuityanditseffectonimpedance[11].IftheimpedanceonaTDR
plotresultsinalowimpedancespikethisisduetothecapacitanceincreasingortheinductance
decreasing in the transmission line. In contrast, if the impedance increases this is due to the
inductanceincreasingorthecapacitancedecreasing.
Z=
T
U
(4.1)
Equation4.2isusedtodeterminethetimetothediscontinuityfromtheSMAendoftheCOTS
adapter(Figure4.1)[19].ThetimecalculatedcorrespondstotheinterfacebetweentheCOTS
adapterandDDMreceptacle(Figure4.2a).Wheretisthetimetothediscontinuity,inseconds,
lW isthelengthofthenthmaterial,inmeters,andϵWY isthedielectricconstantofthenthmaterial.
Forexample,fromTable4.1,theCOTSadapterdielectricconstantis2.1andhasalengthof12.75
mmandtheDDMconnectordielectricconstantis2.4andhasalengthof4.5mm.UsingEquation
4.2,thetime(t)tothediscontinuity(Figure4.2a)is0.17ns.
t=
%Z[ \[]
^
+
%Z_ \_]
^
+ ⋯+
%Za \a
]
^
(4.2)
Figure4.2,showsthemodeloftheconnectorandtheparametricsweepoftheviaoffset.
AsshowninFigure4.2a,thereisadiscontinuityinboththesizeofthestructuresandthematerial
29
propertieswhenconnectingtheCOTSadaptertotheDDMreceptacle.Thediscontinuitycauses
acapacitanceincreaseduetothehigherrelativedielectricconstantof100%in-fillABS,resulting
inanimpedancereduction.OffsettingtheviaintotheDDMconnectorresultsinextrainductance
thatcounteractsthecapacitivediscontinuity.AsmoothertransitionbetweentheCOTSadapter
andtheDDMreceptacleresultsuntiltheinductancebecomestoolarge.Figure4.2bshowsthat
astheviainsetisincreasedtheimpedanceincreases.Theoptimumvaluefortheviaoffsetis0.25
mm.
(a)
(b)
Figure4.2Simulationmodelandresultsvaryingtheviaoffset.
(a)InitialDDMconnectordesign(b)TDRplotofvaryingviaoffset
30
(a)
(b)
Figure4.3Simulationmodelandresultsvaryingthetoptaper.
(a)TaperedDDMconnectordesign(b)TDRPlotwithvaryingthetaperededge
TaperingtheDDMconnector,asshowninFigure4.3a,alsominimizestheabruptchange
intheimpedance.Figure4.3bshowstheparametricsweepofthetaperededge.Thetoptaper
parameteristhedifferencebetweentheouterradiusoftheDDMconnectorandthetopradius
of the taper (Figure 4.3a). Varying the taper doesn’t significantly impact the impedance.
However,thetaperreducesthemechanicalstressesintroducedbythecompressiondesignof
theslip-onconnector.TheCOTSadapterinFigure4.1ahasacompressionfittingontheDDM
connectorend.WhenmatingtheCOTSadapterandtheDDMconnectorthetaperallowsfora
31
smoothertransitiontothelargerradiusoftheDDMconnector.Withoutthetaper,therewould
beasharpedgeattheCOTS-DDMinterfaceandthereisahigherriskofdamagingthetopofthe
connector.
Theanalysisaboveshowsthataviaoffsetof0.25mmandatoptaperof0.4mmarethe
optimumdimensionsfortheconnectorinterface.TheS-parameterperformanceoftheCOTSDDMstructure,withthesemodifications,isshowninFigure4.4.Thedesignresultsinareturn
lossof23.8dBandaninsertionlossof0.2dBat18GHz.
Figure4.4S-parametersofthetaperedconnectordesign
TheDDMconnectorperformswellthrough20GHz.However,theDDMreceptaclemust
transitiontoatransmissionline.TheCOTSadapterisdesignedtomateverticallywithafemale
connector.Inthiscase,thefemaleconnectoristheDDMreceptacle.Mostapplicationsrequire
microstriplinesoranothertypeofplanartransmissionline.Theeffectsoftransitioningfroma
vertical coaxial line to a horizontal microstrip is a challenge. In the next section, the design
challengesandsolutionswillbediscussed.
32
4.3DDMTestFixtureDesign
TheDDMconnector,withoutanytransitionstoaplanartransmissionline,showsgoodS-
parameterperformance.Theconnectormaintainsgreaterthan20dBofreturnlossandlessthan
0.25dBofinsertionlossthrough20GHz(Figure4.4).However,formanyapplicationstheneed
totransitiontoaplanartransmissionlinearises.Inthissection,wewillpresenttheinitialresults
ofaverticaltransitionfromtheDDMconnectortoamicrostripline,optimizationofamicrostrip
taper,andtheoptimizationofanembeddedsemi-circulartransition.
TheteststructureconsistsoftwoDDMconnectorsconnectedbya50-ohmmicrostripline
(Figure4.5).A125μmthick,100%in-fillABS(εr=2.42tanδ=0.006)substrateisusedforthe
substrate.50%in-fillABS(εr=1.6tanδ=0.003)isusedfortheDDMconnector,theeffectof
changingthedielectricisdescribedfurtherinthechapter.Themicrostriplinedimensionsare
calculatedusinglinecalcandareshowninFigure4.5b.TheCOTSadaptermateswiththeDDM
receptaclewiththesignalpincontactingtheinnerconductoroftheDDMreceptacle.Thereisa
missinggroundtoallowtheviatomaketheconnectionwiththemicrostriplineonthebottom
sideofthetestfixturewithoutshortingthesignalpintoground.
(a)
Figure4.5SimulationmodeloftheDDMtestfixture.(a)Cross-sectionoftheDDMconnector
testfixture(b)Bottomviewoftheteststructureshowingthemicrostripsignalline
33
(b)
Figure4.5(Continued)
ThesimulationresultsforthiscircuitareshowninFigure4.6.Thetestfixtureperformance
cutsoffaround4GHz,whichiswellbelowthedesignrequirementof18GHz.Thecauseofthe
degradation in performance is due to the discontinuity between the DDM connector and the
microstripline.Theinductanceatthediscontinuityislargecausingtheimpedancetoincrease
significantly.Anothercauseofthisimpedancemismatchisthemissinggroundplaneabovethe
microstripline,betweentheviaandtheouterconductoroftheDDMconnector;refertothe
zoomedinviewinFigure4.5a.
Figure4.6S-parametersoftheinitialteststructuredesign
34
Afewtechniquescanbeusedtoreducetheinductanceanddiscontinuityatthecoaxial-
microstripinterface.Onesolutionwouldbetoaddamicrostriptaperatthelocationwherethe
coaxialsignallinemeetsthemicrostripline.Figure4.7showsaTDRplotasthetaperwidthis
sweptfrom0.5mmto1.5mm.
Figure4.7TDRplotofchangesintaperwidth
Thetaperdoeshelpwiththeimpedancemismatch,howeverbyitselfthetapercannot
sufficiently reduce the inductance at the discontinuity. To further reduce the inductance, the
groundplaneatthecoaxial-microstripboundaryneedstobebroughtclosertothecoaxialsignal
line. This change will increase the capacitance of the microstrip line, effectively reducing the
impedance.Figure4.8ashowstheTDRresultsofaparametersweepwhichinsetstheground
closertothesignallineoftheDDMconnector.Figure4.8bshowshowthegroundisinsetina
semicircularfashion.Theinsetradiusisthedistancefromtheoutsideoftheconnectortowards
thesignalline.TheTDRshowsthatthe1mminsetradiusoffersthebestimpedancematchalong
theline.However,thisisdeceivingwithoutlookingattheS-parameters.Figure4.9showsthe
insertionlossandreturnlossofthetestfixturewhentheDDMconnectorhasa1mminsetradius.
35
EventhoughtheimpedancematchisreasonabletheaddedstructureiscausingtheE-fieldfrom
theDDMconnectortoabruptlychangetotheE-fieldconfigurationofthemicrostrip.
(a)
(b)
Figure4.8Simulationresultsandmodelvaryingtheinsetradius.
(a)TDRPlotvaryingtheinsetradius(b)3Dmodelincludingtheinsetground
36
Figure4.9S-parametersoftheDDMteststructurewithaninsetradiusof1mm
Even though the characteristic impedance is matched, it is important that the E-field
configuration is also converted effectively between the different transmission line types [20].
Figure4.10showstheE-fieldatthetransition.AtthetransitionpointoftheDDMconnector,the
E-fieldschangetheirorientationby90°totransitiontothemicrostripline.Itcanbeshownthat
iftheinsetgroundisangledupwardsintoasemicircularcone,itwillprovideagroundsection
that will allow the E-fields to gradually transition from a horizontal orientation to a vertical
orientation.
Figure4.10ElectricfieldconfigurationinsidetheDDMconnectorattheDDM-microstrip
transition
37
Theinsetgroundistransformedintoaconewithaheightof0.4mmandaninsetof1mm
asshowninFigure4.11.Aparametersweepofthesemi-circularconeradiusisperformed.The
TDRandS-parametersofthesweepareshowninFigure4.12.
Figure4.11DDMmodelwiththesemi-circularconetransitionaddedwithinthestructure
ATDRanalysisshowsthattheimpedanceismatchedclosestwhentheTopRadiusis1.5
mm.TheS-parameterswithaTopRadiusof1.5mmprovidesmorethan10dBofreturnlossup
to19GHz.
(a)
Figure4.12DDMtestfixturesimulationresultsvaryingthetransitionradius.(a)TDRplot
(b)S-parametersofthesemi-circulartransition(viaoffset=0.25mmtoptaper=0.4mm
microstriptaperwidth=1.5mmtransitionheight=0.4mm)
dashedline–1.75mm,solidline–1.5mm
38
(b)
Figure4.12(Continued)
Keeping the Top Radius parameter constant at 1.5 mm the height parameter can be
swept.TheTDRandS-parameterplotsareshowninFigure4.13.TheS-parametersfor0.5mm,
0.6 mm and 0.7 mm are shown for comparison. It is observed from the TDR plot that the
impedanceincreaseswiththeheightparameter.Thesemi-circularrampequalizestheimpedance
atthediscontinuitytoapproximately50ohms.
(a)
Figure4.13DDMtestfixturesimulationresultsvaryingthetransitionheight.(a)TDRplot(b)Sparametersofthesemi-circulartransition(viaoffset=0.25mmtoptaper=0.4mmmicrostrip
taperwidth=1.5mmtopradius=1.5mm)
Solid–0.5mmDash-Dot–0.6mmDash–0.7mm
39
(b)
Figure4.13(Continued)
The S-parameter data show that the change in height mainly affects mid-band
parameters.Figure4.14showstheE-fieldconfigurationwiththeoptimized(topradius=1.5mm
andtransitionheight=0.6mm)semi-circulartransition.TheS-parametersinFigures4.12band
4.13bshowthatnochangesinthetransitionheightorradiuswillextendtheoperatingfrequency
ofthetestfixture.ThisisduetotheintroductionoftheTE11modewithintheDDMconnector.
Thecut-offfrequencyfortheDDMconnectorcanbecalculatedusingEquation4.3[11],whered
istheouterdiameteroftheinnerconductorandDistheinnerdiameteroftheouterconductor.
Two dielectric constants are possible by changing the in-fill percentage of the ABS within the
DDMconnector,εr=1.6(50%in-fill)andεr=2.4(100%in-fill).Usingtheactualdimensionsofthe
connector(D=6.2mmandd=2mm)thepredictedcut-offfrequenciesassumingεr=1.6andεr
=2.4are18GHzand14.5GHz,respectively.Thesepredictedcut-offfrequenciesdonotmatch
the simulation results, however, because the radius isn’t constant when considering the
embedded transition. The valuefor the innerdiameteroftheouterconductorissmaller.If a
diameterischosenclosetotheaveragevalue,D=4.5mm,thenfc≈23GHzforεr=1.6andfc≈
19GHzforεr=2.4,whichisclosertothesimulatedresultinFigure4.15.
40
Figure4.14ElectricfieldconfigurationoftheDDMconnectorwiththeoptimizedtransition
f^ =
%UP
b cde
\]
(4.3)
One way to increase the operating frequency of the DDM connector is to reduce the
dielectric constant. This will increase the TE11 mode cutoff frequency. An S-parameter
comparisonbetweentheDDMconnectorwithadielectricconstantof2.4and1.6isshownin
Figure4.15.
Figure4.15S-parametersofthestructurewithdifferentdielectricconstants.
Solid–2.4Dash-Dot–1.6
41
4.4DesignConsiderations
Manufacturing technology limitations and processes need to be considered when
designinganystructure,RForotherwise,andDDMisnoexception.Considerationsincludelayer
printorderaswellasverticalconductorprinting.WhendesigningDDMconnectorsthesematters
needtobeaddressed.
Whenconsideringtheprintorderoftheconnectortestfixtureitisquicklyrealizedthat
thecurrentformisnotprintable.Figure4.16helpsdemonstratetheissueswiththeprintability
ofthecurrentdesign.Ifthemicrostriplineisprintedfirsttheconductorwouldbeprintedonthe
heatedprintbed.
Figure4.16DifferentprintingorientationstodemonstratetheissueswithprintingtheDDM
connectortestfixture
If the DDM connector is printed first, then there would be air gaps and it will not be
possible to print the rest of the circuit. A support structure could be used to solve this issue.
However,duetothethinsubstrate,removingthesupportswouldlikelydamagethetestfixture.
TheseissuescanberesolvedbyfirstprintingasmalllayerofABS(Figure4.17).
Figure4.17Modificationrequiredtosuccessfullyprinttheteststructure
42
Adding a 0.1 mm print layer does change the performance of the connector test fixture. The
largest change occurs at 18 GHz with an increase of insertion loss of approximately 0.5 dB.
However,thereturnlossremainsgreaterthan10dBthrough18GHz(Figure4.18).
ViasareasignificantchallengewiththeCB028printingprocess,andverticalprintingis
notcurrentlypossible.Sincetheviascannotbeprinted,theviaoffsetwillberemoved.Fillingthe
viaby-handisn’taccurateenoughtoguaranteea0.25mmviaoffset.
Figure4.18S-parametersoftheteststructurewiththe0.1mmprintlayeradded
TheS-parameterresultsofthetestfixturewithandwithoutaviaoffset,fromFigure4.3,
are shown in Figure 4.19. As mentioned in Section 4.2, the via offset has an impact on the
impedancematchattheCOTS-DDMconnectorinterface.Thechangeintheviaoffsetreduces
thereturnlossatmid-bandfrequenciesandincreasesthereturnlossattheupperfrequencies.
Thechangeintheoffsetlengthhasnoeffectontheinsertionlossofthedesign.
43
Figure4.19S-parametersoftheteststructurewiththemodifiedviaoffset.
Solid–ViaoffsetDash-Dot–Noviaoffset
A final consideration of the design is the mechanical structure needed to prevent the
COTSadapterfromshiftingandpreventingstrains/stressesontheDDMconnector.Toaddress
thisconcern,a100%infillABSstructureisdesignedtoactasareceptaclefortheCOTSadapter.
Thestructureisaddedtothesimulationtoensurethattherearenoperformancedegradations.
Figure4.20showsthefinaldesignandtheS-parameterresultscomparingthesimulationresults
withandwithoutthemechanicalstructure.
(a)
(b)
Figure4.20Modelandsimulationdataoftheteststructurewithandwithoutthemechanical
structure.(a)3Dmodel(b)S-parameters
Solid–MechanicalStructureDash-Dot–NoMechanicalStructure
44
4.5SimulationvsMeasurementResults
Theprevioussectionsdiscussedthegeneraldesignprocessoftheconnectordesign.In
this section, the simulated and measured results will be compared. A cross section of the
fabricateddesignisshowninFigure4.21alongwithapictureofthefabricatedstructurewithout
themechanicalstructureorCOTSadaptersattached.
(a)
(b)
Figure4.213DmodelandfabricatedDDMtestfixture.
(a)Illustrationoffabricatedconnectormodel(b)3Dprintedstructure
TheconnectorfixtureismeasuredwithitsreferenceplanessetattheCOTSadapterinput,
not the DDM connector input. A Keysight PNA-X was used to measure the structure with the
calibrationbeingconductedusingaKeysight85052B3.5mmcalibrationkit.Themeasuredand
simulated results are shown in Figure 4.22. The measurement data doesn’t show very good
45
correlationwiththesimulateddatabeyond14GHz.Thereturnlossthrough14GHzisgreater
than12dBandtheinsertionlossislessthan2.92dB.Theinsertionlossofthemeasurementdata
increasessubstantiallyat15GHz,whichisn’tpredictedinthesimulation.TheCOTSadapterisdeembeddedandthelossesofthemicrostripareremoved.TheresultinginsertionlossoftheDDM
connectoris0.45dB.
Figure4.22MeasuredvssimulatedS-parameterdataoftheteststructure.
Solid–SimulatedDashed–Measured
ThemeasurementsuggeststhatTE11cutoffisbeingshiftedtoalowerfrequency.Thetwo
likely causes were either dielectric constant changes or structural issues, or a combination of
both.Theeasiestissuetocheckisthetransitionwithintheconnector.Thisistheeasiestdueto
thefragilenatureoftheDDMconnectorandtheeasewithwhichitcanbeseparatedfromthe
connectoratthetransition.UponremovaloftheDDMconnectorfromthesubstrate,withthe
transitionexposed,itappearsthattheprintedtransitionwasn’tcompletelycontinuous(Figure
4.23).ThiscouldbeduetothesubsequentABSlayerprintingortheconductorprintingitself.The
CB028gapwasintroducedintotheDDMconnectormodelaswellasanABStransitionasshown
inFigure4.23.IfthetransitionwasnotdamagedtherewouldbenogapsintheCB028semi-
46
circular transition (Figure 4.23a). The gap in Figure 4.23b is only introduced to model the
damagedtransitioninFigure4.23a.
(a)
(b)
(c)
Figure4.23Fabricatedand3Dmodelofdamagedtransition.(a)Photoofthesemicircular
transition(b)3Dmodelofthemodifiedtransitionaccountingfortheconductorgap(c)DDM
embeddedtransitionwithanABStransitionheightadded
Threeparametersareintroduced:GapRadius,GapSize,andABSTransitionHeight.The
ABStransitionisintroducedtosimulatethepossiblechangeinthedielectricduetotheprinting
ofthetransition.Parametersweepsareperformedbynotonlyvaryingtheaboveparametersbut
47
also by including the modified transition in just one and both DDM connectors. A parameter
sweepisperformedwiththemodifiedtransitionincludedinonlyoneoftheDDMconnectors.
TheABStransitionheightisvariedwhileholdingthegapradiusconstant(Figure4.24).Thetwo
gapsizesarealsoshownintheFigure4.24.Figure4.24showsthatvaryingtheABStransition
heightaloneisnotenoughtolowerthecutofffrequency.Changingthegapsizedoesnotshift
theS21resonancesignificantly.However,itdoeshaveasignificantimpactonthelossesatthe
resonance.
Figure4.24S-parametersweepoftheABStransitionheightwithoneconnectordamaged.
GapRadius:2.3mm
Solid–GapSize:0.025mm,Dashed–GapSize:0.05mm
48
Figure4.25S-parametersweepofthegapradiuswithoneconnectordamaged.
ABSTransitionHeight:0.15mm
Solid–GapSize:0.025mm,Dashed–GapSize:0.05mm
Figure4.25isaparametersweepofthegapradiuswhilekeepingtheABStransitionheight
constant. It can be seen that the gap radius contributes greatly to the shift in the resonance
frequency.Changingthegapradiusby0.2mmcausesashiftintheresonancefrequencybymore
than 1 GHz. This significant change is enough evidence to show that the change in the
measurementdatavssimulationdataisduetothepositionoftheconductorgap.However,there
isthepossibilitythattheconductorgapisoccuringinbothoftheDDMconnectors.
TheconductorgapisincludedinbothDDMconnectorsandthesameparametersweeps
areperformed.Asexpectedtheadditionofanotherconductorgapcausesgreaterdisruptionin
thelossesacrossthewholefrequencyband.Figure4.26showsthesimulatedresultsoftheDDM
connectorstructurewhilevaryingtheABStransitionheight.TheABStransitionheightdoesaffect
thelossesbutagainitisnotenoughtochangetheresonancefrequencyalone.
49
Figure4.26S-parametersweepoftheABStransitionheightwithbothconnectorsdamaged.
GapRadius:2.3mm
Solid–GapSize:0.025mm,Dashed–GapSize:0.05mm
Figure4.27S-parametersweepofthegapradiuswithbothconnectorsdamaged.
ABSTransitionHeight:0.15mm
Solid–GapSize:0.025mm,Dashed–GapSize:0.05mm
50
Figure 4.27 shows the simulated performance while varying the gap radius. The S21
resonanceateachgapradiusisconsistentwiththecasewhereonlyonemodifiedDDMconnector
is included. The main difference between one and two modified connectors is the mismatch
acrosstheband.
(a)
(b)
Figure4.28S-parameterdataoftheDDMtestfixturewithadamagedconnector.
(a)onemodifiedtransition(b)twomodifiedtransitions
Solid–Simulated,Dashed–Measured
Thecombinationthatresultsfromtheaboveanalysisareagapsizeof25μmandaradius
of2mm.However,dependingonthecombinationoftheconductorgapwidth,conductorgap
radiusandthethicknessoftheABStransition,similarresultscanbeattained.Figure4.28shows
51
the simulation results of the modified DDM connector transition with only one transition
modified (a) and both transitions modified (b). The simulation with one connector modified
matchesthemeasurementdatacloserthantheDDMconnectorwithbothtransitionsmodified.
4.6TransitionModeling
In the previous sections, the connector structure was modeled using a 3D
electromagneticsolver.However,itisalsousefultocreatealumpedcomponentmodelofthe
connectorstructure.Whenmodelinganycomponent,itisbeneficialtostartatthebasicmodel
togetanapproximationofthebehaviorofthecircuit.Connectorsandtransmissionlinesareno
exception.Figure4.29showsthelumped-elementequivalentcircuitmodelforatransmission
linewithwhichtheconnectorcanbemodeled.AllTEMtransmissionlinescanbemodeledusing
thisconfiguration.Thedeterminationofthelumpedcomponentsiscoveredinmanytextbooks
toinclude[11].TheDDMconnectorfromSection4.5isdeterminedtohaveatransitionalissue
wherethereisagapinthetransitionconductor.Thisgapiscausingtheseconddominantmode,
TE11, to propagate in the desired operating band. Here its shown that by creating a lumped
componentmodel,itispossibletorepresenttheissueinanotherform.
Figure4.29Lumpedcomponentmodelforatransmissionline
TheschematicinFigure4.30showstheoverallcircuitfortheconnectortestfixtureshown
in Figure 4.20. There is good agreement between the 3D model and the lumped component
modelS-parameters(Figure4.31).ThephaseofS21forbothmodelsalsomatchwell.
52
Figure4.30Lumpedcomponentmodeloftheteststructurewithoutthemodifiedtransition
(a)
(b)
Figure4.31Performanceofthelumpedcomponentmodelwithoutamodifiedtransition.
(a)S-parameters(b)Phase
Solid–LumpedComponentModel,Dashed–3DModel
53
ThemodelismodifiedasshowninFigure4.32,toemulatethemeasurementdataofthe
damagedconnector.Thecapacitoracrosstheinductorisaddedtomodeltheconductorgapof
themeasuredconnector.Theshuntcapacitancesarealsovariedinsteadtoaccuratelyrepresent
themodifiedtransition.ComparisonsofthemeasuredandsimulatedS-parametersareshownin
Figure4.33.Theparallelcapacitorisrelatedtothetransitionheightandradius.Forexample,as
the capacitance value is increased this has the same effect as increasing the radius of the
transition. Increasing the transition radius shifts the resonance lower in frequency, as does
increasingthecapacitance.
Figure4.32Lumpedcomponentmodeloftheteststructurewiththemodifiedtransition
(a)
Figure4.33Performanceofthelumpedcomponentmodelwithamodifiedtransition.
(a)S11andS21Magnitude(b)S21Phase
Solid–Simulated,Dashed–Measured
54
(b)
Figure4.33(Continued)
4.7Conclusion
Theconnectordesignprocessrequiresmanystepsandspecialattentionwhenutilizing
additivemanufacturingprocesses.UtilizingtheTDRtoolenablesthedesignertolocatepointsof
adiscontinuitywithinacircuit.Impedanceisanimportantconsideration;however,theelectric
fieldconfigurationisjustascrucialwhendesigningagoodtransition/connector.Manufacturing
considerations need to be addressed when utilizing DDM including: print order and printing
limitations.Electromagneticsimulatorsareinvaluabletoolsthatallowdesignerstotroubleshoot
discrepancies between simulated and measured data. A lumped component model was
developedandprovedusefulinvalidatingphysicaldefectsintheconnectorprototypes.
55
CHAPTER5:DDMTHERMALMANAGEMENT
5.1Introduction
Effective thermal management is paramount for high power and high reliability
applications.Highheatapplicationsutilizemultiplethermalmanagementsystemsincluding:heat
spreaders,heatsinks,andaircooling,tonameafew[21].Thechallengein3Dprintingisthat
thermoplastics are inherently insulators with low thermal conductivities, making it difficult to
effectivelyremoveheatfromasystem.CommonthermoplasticssuchasABSandpolylacticacid
(PLA) have glass transition temperatures below 110°C. There are higher temperature
thermoplasticssuchastheStratasysUTLEMfamily,whichispolyetherimide(PEI).However,the
highertemperaturecouldcauseissuesintheFDMprintingprocess.Insection5.2,wewilldiscuss
afewthermaldefinitionstounderstandtheworkinsection5.4.Insection5.3,wewilldiscuss
thethermalandRFmeasurementsofaKu-bandpoweramplifier(PA).Insection5.4,athermal
modelisdiscussedtodeterminehowtoovercomethechallengesofaPAonthermoplastics.
5.2ThermalBackground
Three heat transfer modes are possible: conduction, convection and radiation. In this
work,onlyconductionandconvectionareconsidered.Conductionisthetransferofheatacross
astationarysolidorfluid.Convectionisthetransferofheatfromasurfacetoamovingmedium
[22].ConductionisimportantintheanalysisofthePAtestfixturebecauseitisthemainthermal
transfermethod.TheheatisgeneratedonthedieofthePAandtravelstothegroundpadofthe
56
package.FromthereitisattachedtoaPCBwherethermalviastransfertheheattothebrass
carrier.Fourier’slawistherateequationusedforheattransfer.Fourier’slawisexpressedas:
q''g = -k
cj
cg
(5.1)
where k is the thermal conductivity, T is the temperature and q is the heat flux [22]. K is an
important material property that describes how well a material can transfer heat. ABS has a
thermalconductivityof0.22W/m·Kwhichmeansitisineffectiveattransferringheatthroughits
structure.BycomparisonH20Eepoxyhasathermalconductivitybetween2.5-29W/m·Kandis
thereforemuchmoreeffectiveatspreadingheat.
Convectionisalsoimportantandisusedinthesimulationsinsection5.4.Itdefinesthe
heattransferfromtheexposedsurfacestotheairsurroundingthetestfixture.Theconvection
coefficient used is defined within Ansys Workbench. The convection coefficient used for the
purposesofthesimulationsisthe“StagnantAir–HorizontalCyl”.Thisoptionprovidedthemost
accurateresultsinthecaseofthepresenteddata.Figure5.1showstheplotoftheconvection
coefficientvstemperature.
Figure5.1ConvectioncoefficientvstemperatureusedinAnsysWorkbench
57
Glass transition temperature is a material property that describes the point at which
materialbeginstofloworbecomesrubberymaterialinsteadofahardandglasslikematerial
[23].Operatingatorexceedingthistemperaturewillcauseissueswithadhesionbetweenlayers
aswellasmechanicalstressessuchaswarping.
5.3ThermalandRFMeasurements
DDMtechnologyisstillrelativelynewtothemicrowavecommunityandalotofquestions
remainunanswered.Increasingfrequencypresentsnewchallengesasdoesincreasingpowerand
temperature.Inthissection,aKubandpoweramplifier’stemperatureandRFperformancewill
beinvestigated.ThedeviceundertestistheQorvoTGA-2527SMpoweramplifier.Atestfixture
designforthePAisprovidedinthedevicedatasheet[24].Thetestfixtureusesan8milRogers
RO4003substratewitharelativedielectricconstantof3.38.Aspreviouslymentioned,100%infillABShasalowerdielectricconstantof2.423.Tomaintainthecharacteristicimpedanceofthe
transmissionlines,thesubstrateheightwasdecreasedto5mils.Theresultingtestboardisshown
inFigure5.2.
Figure5.2ModifiedtestboardusedforanABSsubstrate
58
AstudyoftheRFandthermalperformanceisinvestigatedthroughthevariationofthe
numberofviasusedforthegroundpadofthePA.Thetwoconfigurationsusedhaveeither25or
16thermalvias(Figure5.3).
(a)
(b)
Figure5.3Designswithvaryingvianumbers.(a)25viadesign(b)16viadesign
The components and values used in the bias networks are documented in the device
datasheet[24].TheconnectorsusedforthetestfixturearetheSouthwestMicrowave292-04A5maleSMAendlaunchconnectors.ThefabricatedandassembledtestboardisshowninFigure
5.4.AbrasscarrierisusedtodiffusethehightemperatureawayfromthePA.Theproperbias
conditionsforthePAare:DrainVoltage=6V,Id=650mAandVg=-0.55V.
(a)
(b)
Figure5.4FabricateddesignwithacarrierandSouthwestMicrowaveconnectorsattached.
(a)Topview(b)Bottomview
59
Forthethermaltestingtwodifferenttestareperformed:1)ThePAisterminatedin50
OhmloadsandDCbiased.2)ThePAisplacedinthemeasurementsystemofFigure5.5andtested
withtheapplicationofa14GHzCWsignal.
Figure5.5MeasurementsetupforthePA
FortheDCthermaltesting,thePAisbiasedatdifferentdraincurrentstodeterminethe
typical temperatures for each bias condition. Since the glass transition temperature of ABS is
105°C,it’simportanttodetermineifthesubstrateapproachesthatvalue.Thetemperatureis
measuredfromthetopofthetestfixtureusingtheKeysightU5855ATrueIRThermalImager.
FortemperaturetestingunderDCbiasconditionsthedevicewasbiasedandathermal
imagewastakenafterthetemperaturehadstabilized.BetweeneachbiasconditionthePAis
turnedoffandallowedtoreturntoroomtemperaturebeforethenextbiasconditionisapplied.
Table5.1summarizestheDCbiasresultsforboththe16and25viadesigns.Whiletestingthe16
viadesignitwasdeterminedthatbiascondition5resultsinthemeltingofthe100%infillABS
substrate,atapproximately151°C.Figure5.6showsathermalimageofthedeviceandanimage
ofthedamageddevice.
60
Table5.1Temperaturedataofthetwodesignswhileoperatedovervariousbiasconditions.
Thetemperaturedatawascollectedfromthetopofthechip,withathermalimager.
(a)
(b)
Figure5.6Imagesofthe16viadesignatbiascondition5.
(a)Thermalimage(b)MicroscopeImage
61
There are two possible explanations for why the substrate does not melt when the
thermaltemperaturereaches105°C:1)Thetemperatureismeasuredfromthetop,whichcould
bemeasuringthechanneltemperatureofthedievsthetemperatureatthegroundofthedevice,
and/or2)Thegroundismeltingbutitisnotdetectedduetothechipcoveringthegroundpad
and surrounding area underneath the chip. To test this, the carrier is removed and the
temperature is measured from the bottom of the test fixture. Unfortunately, the heat is not
diffusedawayfromthedeviceefficientlyandasaresultthesubstratemeltedbeforereaching
the150mAdraincurrentcondition.Asaferwaytodeterminethetemperatureunderneaththe
deviceistobiasthePAwiththecarrierattachedandwaitforthetemperaturetostabilize.Then,
turnoffthedeviceandimmediatelymeasurethetemperature.Theresultsofthistechniqueare
coveredlaterinthissection.
Table5.1isusedtodetermineasafebiasconditiontotestthedeviceunderRFdrive,
Table 5.1 is used to determine a safe bias condition. Bias condition 2 is used due to the high
temperatureofthe16viadesign.Figure5.7showsthetemperaturevsoutputpowerofthetwo
designs.The25viadesign,withadraincurrentof100mA,isabout13°Ccoolerthanthe16via
designat16dBmofoutputpower.Sincethe25viadesignhassuperiorthermalperformance,the
deviceisbiasedatmultipledraincurrentsasshowninFigure5.7b.
62
(a)
(b)
Figure5.7Temperaturevsoutputpowerforthetwodesigns.
(a)16viadesign(b)25viaholedesign
Measurementsofthegain,outputpowerandtheassociatedtemperaturesunderthesafe
operating conditions of the PA are performed. Like DC testing, the temperature is allowed to
stabilizebeforeatemperaturemeasurementistaken.UnlikeDCtesting,thetemperatureneeds
to be sampled at each input power of all bias conditions. The device is returned to room
temperaturebetweenbiasconditionsnotbetweeninputpowerchanges.
63
The system setup for the RF and temperature testing is shown in Figure 5.5. The VNA
outputpowerisvariedfrom-15dBmto5dBmwithincrementsof2dBm.A20dBattenuatoris
added to the system to ensure the Mini-Circuits amplifier isn’t driven into compression. The
Anritsupowermeter(ML2438A)iscalibratedtoensureaccuratemeasurementofthepowers
throughthesystem.TheoutputpoweroftheMini-CircuitsamplifierismeasuredwithanAnritsu
powersensor(MA2474A)andAnritsupowermeter.Thesepowersareusedtodeterminethe
inputpowersofthePA.A30dBattenuatorisaddedbetweenthePAandthepowersensorto
preventanydamagetothetestequipment.TheoutputpoweroftheVNAisvariedandtheoutput
powerofthePAismeasured.Thegainandoutputpowervsinputpowerarecalculatedandare
plottedinFigure5.8forbothPAdesigns.ThePAdoesn’tachievetheperformanceprovidedin
thedatasheet.Duetothetemperaturelimitations,thePAcan’tbebiasedtotherecommended
bias conditions. A material with a higher glass transition temperature would extend the
performanceoftheamplifier.
(a)
Figure5.8Outputpowerandgainvsinputpowerforthetwodesigns.
(a)16viadesign(b)25viaholedesign
64
(b)
Figure5.8(Continued)
When biasing the PA to Id = 250 mA the device reaches a maximum temperature of
approximately140°C.Todeterminewhetherit’sthedieitselforthesubstratethatisreaching
thistemperatureadifferenttestisperformed.ThePAispoweredontoabiasconditionofId=250
mA.Afterthetemperaturestabilizesathermalimageisrecorded.Thedeviceisthenpowered
offandanotherthermalimageisrecorded.Ifthetemperatureisnearthetemperatureofthe
devicewhenitison,itcanbeconcludedthatthesubstrateisthattemperature.However,ifthe
temperatureissubstantiallylower,thenitcanbeconcludedthatthedeviceisthecontributorto
thethermalprofileandtheamplifiercanbedriventohighertemperatures.Figure5.9and5.10
showthethermalimageandprofileoftheDUTwhilethedeviceisonanddirectlyafterthedevice
isturnedoff,respectively.Figure5.9bshowsamaximumtemperatureof140°Cwhenthedevice
ison.Figure5.10bshowsamaximumtemperatureofapproximately73°Cdirectlyafterthedevice
is turned off. Figure 5.10a suggests that the temperature of the substrate isn’t reaching the
65
temperature of Figure 5.9a. However, depending on the thermal transient response of the
material,thetemperaturemaychangetooquicklytodetectwiththethermalimagingmethod.
(a)
(b)
Figure5.9ThermalimageandtemperatureprofilewiththePAon.(a)ThermalimageofthePA
testboardwithcarrierat250mA(Id)(b)Temperaturedistributionalongthelineofthethermal
image
(a)
(b)
Figure5.10ThermalimageandtemperatureprofilewiththePAoff.(a)Thermalimageofthe
PAtestboardwithcarrier,turnedoffafterFigure5.9imageistaken(b)Temperature
distributionalongthelineofthethermalimage
ThelimitingfactorofthePAisthetemperatureatwhichitcanoperate.Afewsolutions
tothisproblemarepossible:useamaterialthathasahigherglasstransitiontemperature,design
a heat sink to pull heat away from the top of the device, or possibly electroplate the epoxy
elements.Todeterminethemosteffectiveapproach,asimulationmodelcanbeusedtofindthe
mostfeasibleoption.
66
5.4PowerAmplifierThermalStudy
Inthissection,thethermalperformanceofanamplifiermodelisinvestigated.Theability
tomodelandpredicttheperformanceofanamplifierisinvaluable.Themodelisvalidatedusing
measurementdata.Avalidatedmodelallowsthedesignertodeterminethethermalbehaviorof
the amplifier with changes in the number of vias, the substrate materials, and the use of a
carrier/heatsink.Itisshownthatalltheabovedesignvariablescanbemodeledwithsoftwareto
performtradeoffanalysisofagivendesign.
Thetopologyofagenericquadflatno-leads(QFN)packageisshowninFigure5.11.Inthis
topology,aMMICdieisepoxiedtothepackagegroundingpadwithwirebondsconnectingthe
variouspartsofthechiptothepackagepads.Theentiredieincludingwirebondsisencapsulated
inthepackagecase.ThisQFNamplifiertopologyismountedtoaprintedcircuitboard(PCB)using
solderorepoxy.ThePCBhasalargegroundplaneandthermalviastodiffusetheheatgenerated
onthedieawayfromthechip.Lastly,acarrier/heatsinkisusedtoincreasetheflowofheataway
fromthepackage.
Figure5.11BasictopologyoftheQFNpackagemountedonasubstrateandcarrier
The1WKu-bandPAismodeledalongwitha5milDDMPCBandbrasscarrierinAnsys
Workbench (Figure 5.12). The various materials, thermal conductivities and glass transition
temperatures are listed in Table 5.2. Additionally, various PA structures and dimensions are
67
showninFigure5.13.Itshouldbenotedthatthedimensionsofthepackageandthegroundpad
ofthepackagearetheonlydimensionsgiveninthedatasheet.Therefore,someapproximations
wereusedandtheseapproximationsweretunedtomatchmeasurementdata.
Figure5.12AnsysWorkbenchviewofthesimulationmodel
Table5.2MaterialpropertiesofthevariouscomponentsofthePAtestfixture
Part
Material
Package
Die
Epoxy
Vias
GroundPackage
Substrate
Substrate
Substrate
Polyethylene
GaN
H20E
H20E
Tin
100%In-fillABS
StratasysULTEM9085
StratasysULTEM1010
68
ThermalConductivity
(W/m·K)
0.4
260
2.5
2.5
64
0.22
≈0.22
≈0.22
GlassTransition
Temperature(°C)
-
-
>80
>80
-
105
186
215
Figure5.13QFNstackupandtheassociateddimensions
VariousboundaryconditionsareusedtomodelthePAtestfixture.A1.5Wthermalsource
isappliedtothetopfaceofthedie.Thesourcepowerisdeterminedbythedraincurrentand
voltage. All other exposed surfaces are assigned a convection boundary. Each touching
component is assigned a “bonded” interface. This boundary does not assume any thermal
resistance.Allheatwilltravelacrosscomponentinterfacesunperturbed,inthenormaldirection.
Thesimulationresultsofthe25and16viadesignareshowninFigure5.14and5.15,respectively.
Theresultsarecomparedtothephysicalresultsforbiascondition5inTable5.1.Themaximum
temperatureofthe25viaholemodelisclosetothephysicalmodel.However,the16viadesign
variesby32°Cfromsimulationtomeasurement.Afewcausesthatcanexplainthisvariationare
discussednext.
69
Figure5.1425viaholedesignwitha1.5Wheatsourceappliedtothetopfaceofthedie.
Temperaturerange:78.409°C–103.51°C
Figure5.1516viaholedesignwitha1.5Wheatsourceappliedtothetopfaceofthedie.
Temperaturerange:78.898°C–115.26°C
Thedifferencesbetweenthe16viadesignmeasurementandmodeledperformancecan
be explained by studying the effects of epoxy area and conductivity. When assembling QFN
packages,theamountofepoxyusedandwhetherthechipwaspressedflatwilldeterminethe
footprintoftheepoxy(Figure5.16).Thecuringtemperatureandtimeusedfortheepoxywill
affecttheconductivityoftheepoxy.Asaresult,thethermaldiffusionawayfromthechipwillbe
reduced,increasingtheoveralltemperature.Twosimulationswillbeperformedtodemonstrate
theeffectsthattheepoxyareacoverageandconductivityhaveonthethermalpropertiesofthe
PA.
70
Figure5.16PApadlayoutwithvaryingepoxyareas
AnepoxyinterfaceisaddedbetweenthePAgroundpadandthePCBthermalpad.The
widthandheightoftheepoxyinterfacewillbevariedtoshowthetemperaturechangescaused
byeachconfiguration.Thelengthsandwidthsareequalvaryingfrom1mmto2mmat0.5mm
steps.Figures5.17-5.19showthesimulationresultsofthe16viadesignwith4mm2,2.25mm2
and1mm2areas,respectively.Asthesizedecreasesthemaximumtemperatureincreases.While
thetemperaturedoesincreasethemaximumtemperature,thetemperatureattheedgesofthe
PAfootprintisn’tlargeenoughtoresultintheABSmeltingasexperiencedexperimentally.The
areadoesstillcontributetothetemperatureincrease.However,theareaalonecannotbethe
causeofthetemperaturedifferencebetweenthemodelandmeasurementdata.
71
Figure5.1716viaholedesignwith1.5Wappliedtothetopfaceofthedieanda4mm2epoxy
area.Temperaturerange:78.823°C–117.62°C
Figure5.1816viaholedesignwith1.5Wappliedtothetopfaceofthedieanda2.25mm2
epoxyarea.Temperaturerange:78.68°C–125.99°C
Figure5.1916viaholedesignwith1.5Wappliedtothetopfaceofthedieanda1mm2epoxy
area.Temperaturerange:78.199°C–159.46°C
72
Another cause of the model vs measurement differences is the changes in the epoxy
conductivity.Dependingonthecuretemperature,time,andnumberofcyclestheconductivity
oftheepoxycanchange.AccordingtoEpotekH20Edatasheet,therearetwodifferentthermal
conductivities,2.5W/mKand29W/mK[6].ThelowervalueismeasuredusingtheLaserFlash
method.Wherethelargervalueisbasedonthermalresistancedata[6].Threevaluesarechosen
betweenthisrangetodeterminetheeffectofconductivityonthethermalperformanceandwhat
conductivitymaycontributetodifferencesofthemodeltomeasurement.Figures5.20-5.22show
the simulation results of the PA with conductivities of 29, 15, and 5 W/mK, respectively. The
simulations show that this may be a more plausible explanation for the measurement
differences.Although,itmaybeacombinationofboththeareaoftheappliedepoxyandthe
conductivityduetocuringconfigurations.Basedonthisanalysis,thethermalconductivityused
forH20Eis29W/mKandtheepoxyareais10.9mm2(areaofQFNgroundpad).
Figure5.2016viaholedesignwith1.5Wappliedtothetopfaceofthedieandathermal
conductivityof29W/mK.Temperaturerange:78.895°C–115.17°C
73
Figure5.2116viaholedesignwith1.5Wappliedtothetopfaceofthedieandathermal
conductivityof15W/mK.Temperaturerange:70.644°C–125.11°C
Figure5.2216viaholedesignwith1.5Wappliedtothetopfaceofthedieandathermal
conductivityof5W/mK.Temperaturerange:53.345°C–148.67°C
Todemonstratethe25viamodelaccuracy,Figure5.23showsthemodeledvsmeasured
thermalperformanceovermultiplepowerdissipationranges.Theplotshowsthatoverthebias
conditions of Table 5.1, the model predicts the performance. The data validates the thermal
modelandprovidescertaintyintheaccuracyofthe16viamodel,althoughthedatadoesnot
matchexactly.
74
Figure5.23Maximumtemperaturevspowerdissipatedforthe25viadesign
With the model verified, it can be used to determine a substrate suitable for high
temperatureoperation.Thepowerdissipationisincreasedto3.5WtosimulatethePAworking
attheproperbiasconditionandunderRFdrive.Figures5.24and5.25showthesimulationresults
of the 25 and 16 via designs, respectively. Table 5.2 shows the material properties of two
differenttypesofULTEMmaterials.BothmaterialscanbeusedinthefutureforthePAassuming
thedesignuses25viasandacarrier.Figure5.26showsthesimulationresultsofthe25viadesign,
with1.5Wofpowerdissipationwithoutacarrier.Figure5.26,showsthatacarrierneedstobe
used,whetherit’smetaloranothermaterialwithahighthermalconductivity.
75
Figure5.2425viadesignwith3.5Wappliedtothetopfaceofthedie.
Temperaturerange:125.44°C–183.05°C
Figure5.2516viadesignwith3.5Wappliedtothetopfaceofthedie.
Temperaturerange:124.04°C–206.27°C
Figure5.2625viadesignwith1.5Wappliedtothetopfaceofthediewithoutacarrier.
Temperaturerange:56.779°C–371.18°C
76
5.5Conclusion
AsthetrendofhigherpowerandsmallerdevicesgrowsDDMtechnologyneedstoadapt
to accommodate these changes. The limiting factor of power amplifiers on DDM PCBs is the
temperature that the thermoplastic can handle. The RF performance suffered due to the
temperaturelimitation.However,it’spossiblethatifadifferentthermoplasticisusedthatthe
PAwouldbeabletobebiasedatitspropervoltageandcurrentlevel.ThemodelcreatedinAnsys
Workbench allows the designer to determine whether a specific substrate or carrier/heatsink
configurationwillimprovethethermalperformance.
77
CHAPTER6:CONCLUSION
6.1Summary
Insummary,DDMtechnologyshowspromiseintheRFandmicrowavesfield.Theability
tocreatehighlycustomizedlowcostRFcircuitsisinvaluable.Chapter3showsthattheoverall
size of the DDM hybrid coupler is slightly larger than traditional PCB coupler. However, if a
thermoplasticwithahigherrelativedielectricconstantisusedthesizedifferencewouldn’tbeas
substantial.TheperformancebetweenthePCBandDDMcouplersweresimilar,showingpromise
for3DprintedRFcomponentsat2.45GHz.
Chapter 4 covered the design, measurement and modeling of the DDM Ku band
connector.TheabilitytotunetheconnectorusingtheTDRtoolisinvaluable.Itprovidesinsight
intowhatishappeningwithintheconnectoratdiscontinuities.TheDDMconnectorhaslessthan
0.45dBofinsertionlossupto14GHzandlessthan10dBofreturnlossupto15GHz.Eventhough
theconnectorshowedissueswiththetransition,theperformanceisstillreasonableforafirst
iterationdesign.
Chapter5demonstratedthethermalperformanceofthePAtestfixturewithavarying
number of vias. A thermal model was created in Ansys Workbench which demonstrates
agreementbetweensimulatedandmeasurementdata.Theabilitytousethismodeltoswitch
materialtypesisinvaluableintheevaluationoffuturetests.Workstillneedstobedoneinthe
78
high-poweramplifierfieldof3Dprinting.But,theworkofChapter5providesafoundationfor
futureresearch.
6.2RecommendationsforFutureWorks
FollowonworkintheareasofChapters4and5willbebeneficialtothefutureofDDM.A
majorissuethataroseintheconnectormanufacturingistheprintingoftheembeddedtransition
in Chapter 4. Spraying conductive ink would solve a lot of the issues with the transition. A
different connector or orientation would also be good suggestions. A SMP or SMP-M would
probablybeagoodalternative.SMP-Mhasanupperoperatingfrequencyof40GHzwhichwould
push 3D printing to higher frequencies. Finally, end-launch designs should be considered for
futurework.
Chapter5futureworkwouldincludefurtherdevelopmentofathermalmodelforthePA
device.Validatingthemodeltomeasurementdataformultiplethermoplastics.Themodelcould
be further developed by a mechanical engineer to provide a modeling process for chips of
differentpowersandmaterialtechnologies.
79
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
C.W.Hull,"Apparatusforproductionofthree-dimensionalobjectsbystereolithography,"
1986.Available:https://www.google.com/patents/US4575330.
S. S. Crump, "Apparatus and method for creating three-dimensional objects," 1992.
Available:https://www.google.com/patents/US5121329.
nScrypt.
(2016,
May
31).
nFD
Specification
Sheet.
https://www.nscrypt.com/wp-content/uploads/2017/02/2016-nFD.pdf
Available:
nScrypt. (2016, May 31). SmartPump Specification Sheet. Available:
https://www.nscrypt.com/wp-content/uploads/2017/02/2016-SmartPump-Gen2.pdf
DuPont.(2013,May31).DuPontCB028SilverConductorTechnicalDataSheet.Available:
http://www.dupont.com/content/dam/dupont/products-and-services/electronic-andelectrical-materials/documents/prodlib/CB028.pdf
E. Technology. (2015, May 31). EPO-TEK H20E Technical Data Sheet. Available:
http://www.epotek.com/site/administrator/components/com_products/assets/files/St
yle_Uploads/H20E.pdf
F.Calignanoetal.,"OverviewonAdditiveManufacturingTechnologies,"Proceedingsof
theIEEE,vol.105,no.4,pp.593-612,2017.
nScrypt. (2016, May 31). 3Dn Brochure. Available: https://www.nscrypt.com/wpcontent/uploads/2017/02/2016-3Dn-Brochure.pdf
S.A.Maas,TheRFandMicrowaveCircuitDesignCookbook.ArtechHouse,1998.
R.Mongia,I.J.Bahl,P.Bhartia,andS.J.Hong,RFandMicrowaveCoupled-lineCircuits.
ArtechHouse,2007.
D.M.Pozar,MicrowaveEngineering,4thEdition.Wiley,2011.
R. Corporation. (2016, June 1). RO4725JXR RO4730JXR & RO4730G3 Antenna Grade
Laminates. Available: http://www.rogerscorp.com/documents/1414/acs/RO4725JXRRO4730JXR-RO4730G3-Antenna-Grade-Laminates-Data-Sheet.pdf
80
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
R.Corporation.(2017,June1).RO4000SeriesHighFrequencyCircuitMaterials.Available:
http://www.rogerscorp.com/documents/726/acs/RO4000-LaminatesData-sheet.pdf
N.Arnal,"Astudyon2.45GHzbandpassfiltersfabricatedwithadditivemanufacturing,"
DepartmentElect.Eng,UniversityofSouthFlorida,Tampa,FL,2015.
T.P.Ketterletal.,"A2.45GHzPhasedArrayAntennaUnitCellFabricatedUsing3-DMultiLayer Direct Digital Manufacturing," IEEE Transactions on Microwave Theory and
Techniques,vol.63,no.12,pp.4382-4394,2015.
J. Stratton, "A study of direct digital manufactured RF/microwave packaging,"
DepartmentElect.Eng,UniversityofSouthFlorida,Tampa,FL,2015.
Y.Vega,"AstudyofDigitalRFPhaseShiftersFabricatedwithAdditiveManufacturing,"
DepartmentElect.Eng,UniversityofSouthFlorida,Tampa,FL,2015.
K.RF.(2014,May29).KAD178244.Available:http://konnectrf.com/core/media/
media.nl?id=669029&c=1148964&h=888d20c0eede70b43468
S.Microwave.(2009,June2).UtilizingTimeDoman(TDR)TestMethodsforMaximizing
BoardPerformance.Available:http://mpd.southwestmicrowave.com/showImage.php
?image=1124&name=Utilizing%20Time%20Domain%20(TDR)%20Test%20Methods%20F
or%20Maximizing%20Microwave%20Board%20Performance
E.Holzman,EssentialsofRFandMicrowaveGrounding.ArtechHouse,2006.
C. Bailey, "Thermal Management Technologies for Electronic Packaging: Current
Capabilities and Future Challenges for Modelling Tools," in 2008 10th Electronics
PackagingTechnologyConference,2008,pp.527-532.
F. P. Incropera, F. P. F. o. h. Incropera, and t. mass, Fundamentals of heat and mass
transfer,6thed./FrankP.Incropera...[etal.].ed.Hoboken,NJ:JohnWiley,2007.
E.Technology.(2012,May29).TechTip23.Available:http://www.epotek.com/site/
files/Techtips/pdfs/tip23.pdf
T. Semiconductor. (2015, May 29). TGA2527-SM Ku-Band Power Amplifier. Available:
http://www.qorvo.com/products/d/da004071
81
APPENDIXA:SIMULATIONANDASSEMBLY
A.1Introduction
Throughout the research there was multiple lessons learned when pertaining to
simulations and assembly of DDM parts. From meshing options to the do’s and do not’s of
assemblingwithepoxy.SectionA.2willcoverthesimulationissuethatwasexperiencedinthe
research. Section A.3 covers assembly issues and techniques used throughout the work. This
appendixiswrittentominimizetheissuesthatfuturestudentsmayrunintoduringtheirresearch
inDDMandRFdesign.
A.2SimulationIssues
The major issue that was experienced with simulations was with ADS Momentum
settings. A few settings can be adjusted to ensure that students do not make mistakes when
trying to match simulation and measurement data. The first option (Figure A.1) within the
EMSetupofADSisthe“Simplifythelayout”option.Thisoptionwillresultinthelayoutsmesh
beingreduced.However,indoingso,themeshwillnotaccuratelyrepresentthecircuit(Figure
A.2).Thecellsperwavelengthissetto60andthemeshingfrequencyis3GHz.Thesesettingsare
morethansufficienttosolvethecouplercircuit.However,sincethe“Simplifythelayout”option
isselecteditoverridesthecellsperwavelengtherroneously.Thesamecircuitandoptionsare
usedandthemeshregeneratedwithoutthe“Simplifythelayout”optionanditisseenthatthe
mesh is accurately following the geometry (Figure A.3). This may seem like common sense,
82
howeverwhenusingtheEMCo-simulationoptionthemeshisn’tshown.Toavoidissues,it’sbest
togeneratethemeshpriortorunninganysimulationstoensureanaccuratemeshisused.
FigureA.1EMsetupoptionstoavoidsimplifyingthelayoutmesh
FigureA.2Meshresultingfromthe“Simplifythelayout”option
83
FigureA.3Meshresultingfromnotselectingthe“Simplifythelayout”option
The other options didn’t cause errors, they are just recommendations. The cells per
wavelengthisobviouslyimportantbecauseifit’stwolowthesameresultwilloccurasdidin
FigureA.2.Anothersettingistheedgemeshoption.Thisprovidesmeshspacingbetweenthe
edgesandtheinnerconductor.Itseemstoprovideamoreuniformmesh.Thefinaloption,which
isunchecked,isthe“Meshreduction”option.Although,itdidn’tresultinanerrorinmycircuit
usethisoptionwithcaution.FigureA.4showstheeffectofusingtheMeshreduction.
FigureA.4Meshresultingfromselectingthe“Meshreduction”option
84
A.3AssemblyandFabricationTechniques
Followingafewtechniquesinassemblycanimprovetheend-productofyourdesign.The
firsttipisusedwhenusingamillingmachinetofabricateadesign.Insteadofthemillingmachine
removing the excess conductor from a laminate, it is nicer to just route the traces of the
microstriplinesanduseanExactoknifetopeeltheexcessconductoroff.Thisisn’tatechniqueI
cameupwith,thisissomethingtaughttomebyYanielVega.Thistechniquewillimprovethe
overalluniformityofthesubstrateandminimizeanyroughsurfacesaroundthemicrostriplines.
AssemblywithH20Esometimesisquiteachallenge.Havingassembledover1000parts
myself,Iknowhowtediousthetaskcanbe.Whenmixingthedifferentpartsoftheepoxyitis
extremely important that the epoxy is mixed well. If it is not mixed properly, connectors
especially,willeasilyseparatefromthesubstrateandtransmissionlines.Anotherconsequence
of not mixing properly is that the H20E seems to crack a lot easier. This cracking will cause
connectors to become intermitted and ultimately cause issues with measurement data. Cure
timeandthenumberofcyclesinandoutoftheovenisalsoimportant.Withtheglasstransition
temperatureofABSbeing105°C,thetemperatureusedtocureH20Ewas87°C.TechnicallyABS
isOKat90°C,butasFigureA.5shows,theadhesionbetweenlayersisoftencompromised.Tryto
reduce the number of cure cycles for the part, however this is sometimes unavoidable. If a
mistakeismadewhenmountingalumpedcomponentorconnector(e.g.shortingatransmission
linegap),putthepartintheovenfor1hrat87°Corfinishtheothercomponentsandcureforthe
fullduration.RemovethemistakeAFTERtheparthasbeencured.Thiswillsavealotoftimeand
trouble.Itisaloteasiertoremovethehardenedepoxythattotrytoremovetheuncuredepoxy
withchemicalsorbywipingitaway.
85
FigureA.5TopABSlayerseparatingfromthegroundplane
WhenassemblingQFNpackagestwotechniquesshouldbeconsidered.First,fillthevia
holeswithH20EpriortomountingtheQFN.Thiswillensureallviasareconnectedtoground
providingathermalpathforheatdissipation.CuretheviasbeforeattachingtheQFNpackage.If
the vias aren’t cured without the QFN attached the H20E may not fully cure reducing the
conductivity.Curingtheviasseparatelyalsoensuresthattheepoxyappliedtothegroundpad
will be used for the attachment of the QFN device not filling the vias, possibly causing a
substandardconnection.Second,attachtheQFNpackageandcuretheepoxyagain.Ifthisstep
isn’t performed, it will be extremely hard to make the pad connections without moving the
packagearound.Ensurethattheepoxyisspreadinasomewhatthinlayer.Theepoxyspreading
needs to be limited when pressing the QFN into place, to avoid shorting the ground pad and
signallines.EnsurethattheQFNpackageissittingasflataspossible.Itisn’tacceptabletohave
itsittingoffhighoffthesubstrate.Thiswilldecreasethethermaldiffusionawayfromthechip
andlikelyworsentheRFperformance.Lastly,maketheconnectionstothepads.Again,itisOKif
twopadsshorttogether.Justbakethecircuitandremovetheshortaftertheepoxyiscured.
AnotheradvantageofassemblingtheQFNinmultiplestepsisthatshortsbetweentheground
pad and signal traces can be checked prior to connecting the pads of the package to their
respectivetraces.
86
APPENDIXB:DDMCONNECTORPRINTING
B.13DPrintingProcedure
Asingle0.1μmbaselayerconsistingof100%in-fillABSlayerisprintedwiththefirstlayer
being50μmandtwosubsequentlayersbeing25μmeach.Thesecondlayerconsistsofa25μm
CB028microstriplineismicro-dispensedanddried.AmixtureofABSandacetoneisusedtocover
thebaselayerandmicrostripline.Thismixtureisusedtoimprovetheadhesionbetweenlayers
containingmicro-dispensedconductorsandthenextABSlayer.The125μmsubstratelayeris
printed next (100% in-fill ABS) followed by the ground plane, which is micro-dispensed and
allowedtodry.ThetwoDDMconnectors(50%in-fillABS)areoutlinedusing100%in-fillABS.The
semi-circulartransitionisprintedinastaircasefashionwithaz-axisresolutionof25μm.The
CB028pasteismicro-dispenseoverthetransitionanddried.Whenthetransition(100%in-fill
ABS)isfinished,the50%in-filllayersarecontinuedin50μmstepstofinishtheremainderofthe
connector.Theconductiveinkontheouterportionsoftheconnectorsandviaholesarepost
processed by-hand. The mechanical structures are printed separately from the connector
structureandattachedusingadhesiveformechanicalstrength.ThenScrypt3Dnprintercanprint
themechanicalstructuresinsitu.Thedecisiontoprintthestructuresseparatelywasduetothe
postprocessingrequiredtoensuretheconnectorsfitwascorrect.
87
Документ
Категория
Без категории
Просмотров
0
Размер файла
29 892 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа